# Spend Control
Source: https://getskydive.com/docs/account/cost-tips

Set monthly budgets, then spend less: model selection, agent design, and event-driven runs

> [!NOTE]
>
> **Skydive does not mark up model usage.** What model providers charge is exactly what draws your credit balance. No platform fee on top, no hidden premium on tokens. Every dollar goes directly to the work your agents do.

Cap what the workspace can spend, then spend less on the work itself. Budgets live on the Spend controls tab. Model choice, agent design, and when agents wake up do the rest.

## Spend controls

Owners and admins set monthly budgets on the **Spend controls** tab of [Usage & billing](https://www.skydive.com/settings/billing). Members who are allowed to see their own spend see a limit and a running total on [Usage](/docs/account/usage) instead.

* **Monthly budget.** Optional workspace ceiling for the billing period. When spend crosses it, new agent runs pause until you raise or remove the budget, or the period resets. Runs already in flight are not cut off mid-task. Minimum \$1, maximum \$1,000,000.
* **Warn admins at.** Optional heads-up email before the budget is reached (50%, 75%, 80%, or 90%). If you leave this off, the first email is the one that says the budget ran out.
* **Default agent limit.** Optional monthly cap that applies to every agent unless you set a different one below. Agents with no extra cap spend against the workspace budget.
* **Additional limits.** Caps on top of the workspace budget for individual agents, members, teams, or people within a team. A row that says **No additional limit** is not uncapped. It still spends against the workspace budget. You can pick a preset or type a custom dollar amount.
* **What members can do.** Choose who can see their own spend (everyone, admins, or owners only) and who can change an agent's limit (anyone who can edit the agent, admins, or owners only).

Limits reset when the billing period resets, not necessarily on the 1st of the month. Auto-refill lives on the Billing tab. See [Billing](/docs/account/pricing#controlling-spend).

## Watch the Usage dashboard

The [Usage page](/docs/account/usage) breaks spend down per agent and over time. The common pattern: one agent with an overfit model or a too-frequent cron is driving a surprising share of total spend. Find it, look at what it is doing, and fix the model or the schedule.

## Pick the right model for the job

Model tokens are the biggest lever. The difference between a flash model and a flagship can be 100x in price per token, and most agent tasks do not need a flagship.

A quick rule: match the model to the hardest thing the agent does regularly, not the hardest thing it could ever do. An agent that summarizes support tickets, routes leads, or monitors a feed is doing structured work on known inputs. It does not need a reasoning model with a \$50/M output rate.

### What the tiers cost

Rates are USD per 1M tokens (input / output), passed through at cost.

| Model | In | Out | Best for |
| - | - | - | - |
| **Qwen 3.7 Flash** | \$0.03 | \$0.13 | Very high volume, simple classification, structured extraction |
| **DeepSeek V4 Flash** | \$0.13 | \$0.26 | Fast loops, drafting, routing |
| **Gemini 3 Flash** | \$0.50 | \$3 | Light reasoning with Google ecosystem |
| **Claude Haiku 4.5** | \$1 | \$5 | Trusted Anthropic option at a low rate |
| **Grok Build 0.1** | \$1 | \$2 | Coding tasks on a budget |
| **DeepSeek V4 Pro** | \$1.74 | \$3.48 | Strong general capability per dollar |
| **Claude Sonnet 4.6** | \$3 | \$15 | Reliable everyday default, broad tasks |
| **GPT 5.6 Terra** | \$2 | \$12 | Balanced GPT for routine agent work |
| **Claude Opus 4.8** | \$5 | \$25 | Hard judgment, ambiguous briefs, careful output |
| **Claude Fable 5** | \$10 | \$50 | Anthropic's top reasoning model |

The full catalog with context sizes and notes is at [Models](/docs/account/models-catalog).

### Turn thinking down

Reasoning-capable models expose a thinking level. Higher thinking burns more tokens. For most everyday tasks, low or off is fine. Save high thinking for genuinely hard problems, not for every run.

### Lock in a default, escalate deliberately

Set a workspace default model that is cheap and capable enough for routine work. Switch individual agents up when their job genuinely warrants it. That way a new agent someone spins up does not silently land on a flagship.

In **Settings, Models**, you can also hide models from pickers entirely. If half your team does not need access to the top-tier options, remove the temptation.

## More agents, lower cost

This sounds backwards, but it is true in practice. A single general-purpose agent handling ten different kinds of work carries all ten contexts at once. A long, mixed context means more tokens on every turn, regardless of which task is actually happening.

Split that agent into focused specialists and each one runs a shorter, tighter context. The Intercom agent knows support. The GitHub agent knows code. They do not need to hold each other's world in memory on every turn.

Focused agents also let you match models precisely. Your support triage agent can run on Haiku or a flash model. Your code review agent gets Sonnet. Your weekly executive summary agent is the only one that needs a flagship, and it runs once a week. You are not paying Opus rates for the agent that routes tickets.

The [Usage dashboard](/docs/account/usage) makes this visible. Per-agent spend tells you which agent is running heavy. That is usually the one doing too many jobs.

## Wake on event, not on a schedule

A cron-polling agent wakes up, loads its context, calls the API, reasons over the result, and does nothing if nothing arrived. You paid for the full run anyway.

A webhook-triggered agent sleeps at zero cost until the service sends an event. It wakes once per real event, does the work, and stops. For most support, sales, and notification workflows, that is a significant reduction in idle spend with no change in what gets done.

Use a scheduled cron when the trigger is genuinely time-based: a weekly digest, a daily standup, a monthly report. Use a webhook when the trigger is an event: a new ticket, a merged PR, a payment landing. Any service that can POST to a URL works.

Ask your agent to set up a webhook receiver and it will handle the endpoint, signature verification, and deduplication automatically using its built-in [webhook-receiver skill](/docs/agents/skills).

> **[Model catalog](/docs/account/models-catalog)**
>
> Every live model with rates, context sizes, and notes.

> **[Usage dashboard](/docs/account/usage)**
>
> Per-agent spend, over time, against your balance.

> **[Webhooks](/docs/capabilities/webserver#receiving-webhooks)**
>
> How your agent receives inbound events from external services.

> **[Pricing](/docs/account/pricing)**
>
> Plans, credit, spend limits, and auto-refill.


# Model Settings
Source: https://getskydive.com/docs/account/model-settings

Choose from 300+ models and control how providers use your data

You can turn on additional controls for your workspace on how models use your data when you run your agents. These limit the model choices available to you.

## No training

Only use providers whose agreements or API terms prohibit training on your data. Providers may still retain data for other purposes, such as abuse monitoring. Skydive will never train its own models to improve agents on your data.

## Zero data retention (ZDR)

Only use providers configured to retain none of your prompts or responses, including for abuse monitoring. This is the strictest setting and offers the fewest models.

## Google integrations

Agents connected to Google services automatically use models that meet Google's Limited Use requirements, including no training on your Google data. As a result, model choice is limited further.

## Which models can I use?

Skydive offers [300+ models](/docs/account/models-catalog) to run your agents, including frontier models from OpenAI, Anthropic, xAI, and Google and leading open-source models.

You can train your agent on any model and it retains its memories when you switch. Over time, you can lower costs by moving it to an open-source model. In some cases, the model costs one-tenth as much.

[Skydive Glide](/docs/account/models#skydive-glide) is our default model router. It chooses the best model for your task and balances cost and quality without you needing to switch.

Controls like No Training and ZDR apply at the routing layer. Each model can be served by multiple providers with different data terms, so availability depends on the provider, not just the model's name. Open-source models, including those behind Glide, can qualify.

These controls apply to every request and fallback. If no eligible provider is available, the request fails rather than using one that violates your settings. Check your agent's model picker for the models available to you.

These controls govern model-provider handling of your data, not your conversation history in Skydive.


# Models
Source: https://getskydive.com/docs/account/models

Choose which AI models power your agents, and keep spend under control

Skydive separates the agent from the model that runs it. You can change the model at any time (choose from 300+ frontier and open source models). You pay **0% markup on provider API rates**.

For no-training and zero data retention settings, see [Model Settings](/docs/account/model-settings).

## Choosing a model

> **Change the model on an agent**
>
> Open the agent and use the model picker to switch which model it runs.

> **Workspace default model**
>
> Set a default model for new agents in **Settings, Models**.

> **Search any model**
>
> The picker shows seven curated models by default. Search to reach the full catalog.

> **[Full model catalog](/docs/account/models-catalog)**
>
> Every live model with notes, context size, and at-cost rates.

> [!NOTE]
>
> New agents start on **Skydive Glide** by default, a fast, cost-effective
> model Skydive manages on your behalf. You can switch any agent to a
> different model at any time. See [Skydive Glide](#skydive-glide) below.

> [!NOTE]
>
> Connect an approved Google Workspace app (Gmail, Drive, Calendar, Docs, Sheets, Slides, Contacts, Tasks, Forms) and that agent locks to [trusted models](/docs/agents/trusted-models). Other models stay visible but cannot be selected while the connection is active.

## Change the model on an agent

The model is set **per agent**. Switching updates that agent only. The new model (and thinking level, if you set one) applies on the agent's next run.

1. Open the agent (from chat or the agent profile).
2. Open the **model picker** (the model name by the composer or agent controls).
3. Choose from the seven featured models, or search to reach any model in the catalog.
4. Hover a model to see its intelligence, speed, and cost ratings and pick a thinking level.
5. Confirm the switch.

| Control | What it changes |
| - | - |
| **Model picker, inside a conversation** | Pins to that conversation only. Other threads keep whatever model they were using. |
| **Model picker, on the new-chat screen** | Sets the agent's default for all future conversations. |
| **Settings, Models** | Set the workspace default model for new agents. Existing agents keep their current model. |

When you switch models inside a live conversation thread, the change pins to that thread alone. Your other open conversations and the agent's default are untouched. If you open the picker on the new-chat screen before a conversation starts, the selection becomes the agent's new default instead.

> [!TIP]
>
> Put everyday agents on a balanced model. Switch one agent to a heavier model when the job needs it, instead of changing every agent.

## How the picker works

The picker opens with seven curated models: Glide, Claude Opus 5.5, GPT-6.1 Sol, Grok 4.7, Claude Haiku 5.5, Gemini 4 Argon, and GLM 5.3 Flash. If your agents are running a model outside that set, it shows as an extra row until you switch.

Rest on any model to see its intelligence, speed, and cost ratings (sourced from Artificial Analysis) and click a thinking level directly from the card. Search to reach any model in the full catalog. Models you pick from search save to your set and show in future sessions.

Models that can generate images show a small image icon next to their name, so you can spot them when a task calls for one.

### When your workspace is out of credits

What the picker shows depends on which plan you're on.

On the free plan, [Glide serves as the free fallback](/docs/account/pricing) and the picker trims down to the free models only. The current model is checked and tagged "free", and a **See all models** link opens the full catalog page so you can see what paid plans include. There are no locked rows to scroll past.

On a paid plan with an empty balance, the full list stays visible but non-Glide models are dimmed and can't be selected until you add credits. In the TUI, the status bar shows "needs credits" while a locked model is highlighted.

Once you add credits, the picker returns to its normal state.

## Match the model to the job

| Kind of work | What to reach for |
| - | - |
| Hard reasoning, big builds, careful judgment | A flagship model (Opus-class, top GPT, top Gemini). |
| Everyday agenting | A balanced mid-tier model. This is the right default for most agents. |
| High volume, quick lookups, light drafting | A flash, mini, or haiku-class model, or a free variant when quality is enough. |
| Google-connected agent | A [trusted model](/docs/agents/trusted-models) only (Trusted in the catalog). |

Reasoning-capable models expose a **thinking level**. Turn it up for harder problems; turn it down for speed and lower cost.

## Cost-effective picks

**Model usage is at cost.** Provider rates pass through to your [credit balance](/docs/account/pricing) with no markup on model spend. Costs below are USD per **1M tokens** (input / output), from the live registry.

### Strong defaults

| Model | Provider | In | Out | Why pick it |
| - | - | - | - | - |
| **Claude Sonnet 5.5** | Anthropic | \$2 | \$10 | Everyday default; strong quality without Opus rates |
| **GPT-6.1 Sol** | OpenAI | \$2 | \$10 | Balanced GPT tier for routine agent work |
| **Grok 4.7** | xAI | \$2 | \$6 | Latest Grok; fast reasoning and tool use |
| **Gemini 3.1 Pro Preview** | Google | \$2 | \$12 | Large-context Google option at a mid rate |
| **GLM 5.3** | Zhipu | \$0.08 | \$4.8 | Trusted reasoning model; strong on coding and agent workflows |
| **DeepSeek V4 Pro** | DeepSeek | \$0.41 | \$0.82 | High capability per dollar |

### When you need the flagship

| Model | Provider | In | Out | Why pick it |
| - | - | - | - | - |
| **Claude Opus 5.5** | Anthropic | \$4 | \$20 | Anthropic's top reasoning model; trusted and compliant |
| **Claude Opus 4.8** | Anthropic | \$5 | \$25 | Hardest agent jobs and careful judgment |
| **GPT-6 Astra** | OpenAI | \$10 | \$50 | Top OpenAI option for heavy multi-step work |
| **Kimi K3** | Moonshot | \$0.73 | \$10 | Long coding and reasoning sessions |

### Thrifty and high volume

| Model | Provider | In | Out | Why pick it |
| - | - | - | - | - |
| **Gemini 3.8 Flash** | Google | \$0.75 | \$3.75 | Fast, trusted Gemini with Pro-level agentic capabilities; multimodal |
| **Qwen 3.7 Flash** | Qwen | \$0.03 | \$0.13 | Very low cost with huge context |
| **DeepSeek V4.1 Flash** | DeepSeek | \$0.15 | \$0.60 | Beats V4 Pro on agentic benchmarks at a fraction of the price |
| **Grok Build 0.1** | xAI | \$1 | \$2 | Coding-focused and inexpensive |
| **Claude Haiku 5.5** | Anthropic | \$0.10 | \$0.50 | Trusted Anthropic option that stays cheap |

## Skydive Glide

Glide is Skydive's own model option. It delivers strong everyday performance
across agent work (drafting, research, tool use, multi-step tasks) and it
is the default for new agents when your workspace has not set a different
default.

A few things that make it different from picking a provider model directly:

* **Skydive manages the target.** Glide is an alias that routes to the
  best available model for the job at any given time. Skydive updates the
  target when a better option emerges, so your agents stay current without
  you changing anything.
* **Quality and cost in balance.** Glide sits at a mid-tier price point
  while targeting performance you would expect from a capable everyday
  model. Rates appear in the [model catalog](/docs/account/models-catalog).
* **No upcharge on any model.** Whether you run Glide or pick a model
  yourself, Skydive passes provider rates through at cost. You pay the
  same rate regardless of which model is underneath.
* **Fallback is built in.** If the underlying model is unavailable, Glide
  routes to a backup at the same or lower price, never up to a premium model.

Glide is a solid default for most agents. Switch to a flagship when the
work is genuinely hard: long reasoning chains, careful judgment, ambiguous
briefs. Switch to a flash model when you are running high volume and the
task is light.

## Use your own provider or gateway key

Workspace owners and admins can connect their own API keys for billing,
discounts, compliance, or provider account controls.

1. Open your workspace settings.
2. In **AI providers**, select **Add provider**.
3. Choose a provider or gateway.
4. Enter its API key, then select **Connect**.

Skydive verifies the key before saving it. The connection applies to the
entire workspace.

You can connect a gateway key for **OpenRouter** or **Vercel AI Gateway**.
Skydive routes compatible models through your gateway account. Your gateway
controls billing and any provider routing you configured there.

You can also connect a direct provider key for Anthropic, OpenAI, Google AI
Studio, xAI, or Z.ai. A direct provider key takes priority over a gateway key
for that provider. Other compatible models continue through the gateway.

To stop using a key, open **AI providers** in your workspace settings and
select **Disconnect**.

## Spend tips

* **Default cheap, escalate one agent.** Keep most agents on a mid or flash model. Only the agent doing the hard job needs the flagship.
* **Turn thinking down when you can.** Higher thinking levels burn more of your credit balance on reasoning-capable models.
* **Set a workspace default.** In Settings, Models, pick the default model so new agents start on the right tier without each one needing manual setup.
* **Watch the Usage dashboard.** See which models draw credit under [Usage](/docs/account/usage), and swap offenders to a thriftier pick.
* **Prefer at-cost clarity over guessing.** Rates in the [catalog](/docs/account/models-catalog) are the same pass-through rates billed to your credit balance.

## How cost ties in

The model you choose affects both quality and [cost](/docs/account/pricing). More capable models and higher thinking levels draw more from your credit balance per task.

**Model usage is at cost.** Provider rates pass through with no markup. You pick the model for the job, not to dodge a hidden platform fee on model spend.

> **[Model catalog](/docs/account/models-catalog)**
>
> Full live list with notes and rates.

> **[Pricing](/docs/account/pricing)**
>
> Credit balance, at-cost model usage, and plans.

> **[Usage](/docs/account/usage)**
>
> Breakdown of where credit went.

> **[Trusted models](/docs/agents/trusted-models)**
>
> Google connections and the trusted set.


# Model catalog
Source: https://getskydive.com/docs/account/models-catalog

Every model available to Skydive agents, with rates and short notes

Full list of models Skydive agents can run today (**389** from **47** providers). Rates are **at cost** (USD per 1M tokens, no markup). This page regenerates from the live product registry.

> [!NOTE]
>
> Generated from the live model registry. Last synced **2026-10-28** (UTC). The model picker is source of truth if anything drifts.

## How to read this page

| Column | Meaning |
| - | - |
| **Model** | Name in the model picker |
| **Note** | One-line summary |
| **Context** | How much the agent can read at once |
| **In / Out** | At-cost USD per 1M input / output tokens |
| **Trusted** | OK for [Google-connected](/docs/agents/trusted-models) agents |

Thinking-capable models also expose a **thinking level** in the picker (for example low, medium, high). Higher levels cost more per run.

For how to switch models and save spend, see [Models](/docs/account/models).

## By provider

> **Anthropic (14)**
>
> | Model | Note | Context | In | Out | Trusted |
> | - | - | - | - | - | - |
> | Claude Opus 4.1 | Claude Opus 4.1 is an updated version of Anthropic s flagship model, offering improved performance... | 200K | \$15 | \$75 | Yes |
> | Claude Fable 5 | Claude Fable 5 is a Mythos-class model with safeguards. It can handle long-running, complex, and... | 1M | \$10 | \$50 | |
> | Claude Fable 5.1 | Claude Fable 5.1 improves on Fable 5 across long-running agentic coding, knowledge work, and... | 1M | \$10 | \$50 | |
> | Claude Opus 4.8 (Fast) | Opus 4.8 is a focused upgrade to Opus 4.7 and is Anthropic's best generally available model for... | 1M | \$10 | \$50 | |
> | Claude Opus 5 (Fast) | Claude Opus 5 is the latest model in Anthropic's Opus family and a step-change improvement over... | 1M | \$10 | \$50 | |
> | Claude Opus 5.5 (Fast) | Claude Opus 5.5 is a step-change improvement over Opus 5 for agentic coding, long-running agentic... | 1M | \$8 | \$40 | |
> | Claude Opus 4.5 | Claude Opus 4.5 is Anthropic s latest model in the Opus series, meant for demanding reasoning... | 200K | \$5 | \$25 | Yes |
> | Claude Opus 4.8 | Opus 4.8 is a focused upgrade to Opus 4.7 and is Anthropic's best generally available model for... | 1M | \$5 | \$25 | Yes |
> | Claude Opus 5 | Claude Opus 5 is the latest model in Anthropic's Opus family and a step-change improvement over... | 1M | \$5 | \$25 | Yes |
> | Claude Opus 5.5 | Claude Opus 5.5 is a step-change improvement over Opus 5 for agentic coding, long-running agentic... | 1M | \$4 | \$20 | Yes |
> | Claude Sonnet 5 | Sonnet 5 is an upgrade to Sonnet 4.6, with gains across agentic coding and professional work. It... | 1M | \$2 | \$10 | Yes |
> | Claude Sonnet 5.5 | Claude Sonnet 5.5 is a step-change improvement over Sonnet 5 for well-scoped everyday work like... | 1M | \$2 | \$10 | Yes |
> | Claude Haiku 4.5 | Claude Haiku 4.5 matches Sonnet 4's performance on coding, computer use, and agent tasks at... | 200K | \$1 | \$5 | Yes |
> | Claude Haiku 5.5 | Claude Haiku 5.5 brings adaptive reasoning to Anthropic s low-cost Haiku family for everyday... | 1M | \$0.1 | \$0.5 | Yes |

> **OpenAI (88)**
>
> | Model | Note | Context | In | Out | Trusted |
> | - | - | - | - | - | - |
> | GPT 5.4 Pro | GPT-5.4 Pro uses more compute to think harder and provide consistently better answers. It's... | 1.1M | \$30 | \$180 | Yes |
> | GPT 5.5 Pro | Reasoning model, 1M+ context. | 1.1M | \$30 | \$180 | |
> | GPT-4 | OpenAI's flagship model, GPT-4 is a large-scale multimodal language model capable of solving... | 8K | \$30 | \$60 | Yes |
> | GPT 5.2 | Version of GPT-5.2 that produces smarter and more precise responses. | 400K | \$21 | \$168 | |
> | GPT-6 Astra (Fast) | GPT-6 Astra is OpenAI's most capable model for complex reasoning, coding, computer use, research... | 1.1M | \$20 | \$100 | |
> | o3 Pro | The o-series of models are trained with reinforcement learning to think before they answer and... | 200K | \$20 | \$80 | |
> | GPT-5 pro | GPT-5 pro uses more compute to think harder and provide consistently better answers. Since GPT-5... | 400K | \$15 | \$120 | |
> | o1 | o1 is OpenAI's flagship reasoning model, designed for complex problems that require deep thinking... | 200K | \$15 | \$60 | |
> | GPT 5.5 (Fast) | GPT-5.5 understands what you re trying to do faster and can carry more of the work itself. It... | 1M | \$12.5 | \$75 | |
> | GPT-4 Turbo | The latest GPT-4 Turbo model with vision capabilities. Vision requests can now use JSON mode and... | 128K | \$10 | \$30 | |
> | GPT-6 Astra | GPT-6 Astra is OpenAI's most capable model for complex reasoning, coding, computer use, research... | 1.1M | \$10 | \$50 | Yes |
> | GPT-6 Astra Pro | GPT-6 Astra Pro is the same underlying model as GPT-6 Astra, served with reasoning.mode set to pro... | 1.1M | \$10 | \$50 | Yes |
> | GPT 5.6 Sol (Fast) | GPT-5.6 Sol is the flagship of OpenAI's GPT-5.6 series, its most capable model for long-horizon... | 1.1M | \$8 | \$40 | |
> | GPT 5.4 (Fast) | GPT-5.4 is OpenAI's best general-purpose model, part of the GPT-5 flagship model family. It's... | 1.1M | \$5 | \$30 | |
> | GPT 5.5 | GPT-5.5 understands what you re trying to do faster and can carry more of the work itself. It... | 272K | \$5 | \$30 | Yes |
> | GPT Chat Latest | GPT Chat Latest points to OpenAI's stable API alias chat-latest that always resolves to the latest... | 400K | \$5 | \$30 | |
> | GPT Image 1 | Standard model. | - | \$5 | \$40 | |
> | GPT Image 2 | Standard model. | - | \$5 | \$30 | |
> | GPT-4o (2024-05-13) | GPT-4o ("o" for "omni") is OpenAI's latest AI model, supporting both text and image inputs with... | 128K | \$5 | \$15 | Yes |
> | GPT-4o (Fast) | GPT-4o from OpenAI has broad general knowledge and domain expertise allowing it to follow complex... | 128K | \$4.25 | \$17 | |
> | GPT 5.6 Sol | GPT-5.6 Sol is the flagship of OpenAI's GPT-5.6 series, its most capable model for long-horizon... | 272K | \$4 | \$20 | Yes |
> | GPT 5.6 Terra (Fast) | GPT-5.6 Terra is a balanced GPT-5.6 model for everyday work, with performance comparable to the... | 1.1M | \$4 | \$24 | |
> | GPT-5.6 Sol Pro | GPT-5.6 Sol Pro is the same underlying model as GPT-5.6 Sol, served with reasoning.mode set to pro... | 1.1M | \$4 | \$20 | Yes |
> | GPT-6 Sol (Fast) | GPT-6 Sol is OpenAI's reasoning model for complex coding and agentic workflows. It accepts text... | 1.1M | \$4 | \$20 | |
> | GPT-6.1 Sol (Fast) | GPT-6.1 Sol is OpenAI's reasoning model for complex coding, computer use, and professional work... | 1.1M | \$4 | \$20 | |
> | GPT 5.2 (Fast) | GPT-5.2 is OpenAI's best general-purpose model, part of the GPT-5 flagship model family. It's... | 400K | \$3.5 | \$28 | |
> | GPT 5.3 Codex (Fast) | GPT-5.3-Codex advances both the frontier coding performance of GPT-5.2-Codex and the reasoning and... | 400K | \$3.5 | \$28 | |
> | GPT-4.1 (Fast) | GPT 4.1 is OpenAI's flagship model for complex tasks. It is well suited for problem solving across... | 1M | \$3.5 | \$14 | |
> | o3 (Fast) | OpenAI's o3 is their most powerful reasoning model, setting new state-of-the-art benchmarks in... | 200K | \$3.5 | \$14 | |
> | GPT-3.5 Turbo 16k | This model offers four times the context length of gpt-3.5-turbo, allowing it to support... | 16K | \$3 | \$4 | Yes |
> | GPT 5.1 Thinking (Fast) | An upgraded version of GPT-5 that adapts thinking time more precisely to the question to spend... | 400K | \$2.5 | \$20 | |
> | GPT 5.4 | GPT-5.4 is OpenAI's best general-purpose model, part of the GPT-5 flagship model family. It's... | 272K | \$2.5 | \$15 | Yes |
> | GPT Audio | The gpt-audio model is OpenAI's first generally available audio model. The new snapshot features... | 128K | \$2.5 | \$10 | |
> | GPT Image 1 Mini | Standard model. | - | \$2.5 | \$32 | |
> | GPT-4o | GPT-4o from OpenAI has broad general knowledge and domain expertise allowing it to follow complex... | 128K | \$2.5 | \$10 | Yes |
> | GPT-4o (2024-08-06) | The 2024-08-06 version of GPT-4o offers improved performance in structured outputs, with the... | 128K | \$2.5 | \$10 | Yes |
> | GPT-4o (2024-11-20) | The 2024-11-20 version of GPT-4o offers a leveled-up creative writing ability with more natural... | 128K | \$2.5 | \$10 | |
> | GPT-5 (Fast) | GPT-5 is OpenAI's flagship language model that excels at complex reasoning, broad real-world... | 400K | \$2.5 | \$20 | |
> | GPT 5.6 Terra | GPT-5.6 Terra is a balanced GPT-5.6 model for everyday work, with performance comparable to the... | 272K | \$2 | \$12 | Yes |
> | GPT-4.1 | GPT 4.1 is OpenAI's flagship model for complex tasks. It is well suited for problem solving across... | 1M | \$2 | \$8 | Yes |
> | GPT-5.6 Terra Pro | GPT-5.6 Terra Pro is the same underlying model as GPT-5.6 Terra, served with reasoning.mode set to... | 1.1M | \$2 | \$12 | Yes |
> | GPT-6 Sol | GPT-6 Sol is OpenAI's reasoning model for complex coding and agentic workflows. It accepts text... | 1.1M | \$2 | \$10 | Yes |
> | GPT-6 Sol Pro | GPT-6 Sol Pro is the same underlying model as GPT-6 Sol, served with reasoning.mode set to pro for... | 1.1M | \$2 | \$10 | Yes |
> | GPT-6.1 Sol | GPT-6.1 Sol is OpenAI's reasoning model for complex coding, computer use, and professional work... | 1.1M | \$2 | \$10 | Yes |
> | GPT-6.1 Sol Pro | GPT-6.1 Sol Pro is the same underlying model as GPT-6.1 Sol, served with reasoning.mode set to pro... | 1.1M | \$2 | \$10 | Yes |
> | GPT-6.1 Terra (Fast) | GPT-6.1 Terra is OpenAI s model for visual reasoning, precise computer use, and natural writing... | 1.1M | \$2 | \$10 | |
> | o3 | OpenAI's o3 is their most powerful reasoning model, setting new state-of-the-art benchmarks in... | 200K | \$2 | \$8 | |
> | GPT 5.2 | GPT-5.2 is OpenAI's best general-purpose model, part of the GPT-5 flagship model family. It's... | 400K | \$1.75 | \$14 | Yes |
> | GPT 5.2 Codex | GPT-5.2-Codex is a version of GPT-5.2 further optimized for agentic coding in Codex, including... | 400K | \$1.75 | \$14 | Yes |
> | GPT 5.3 Codex | GPT-5.3-Codex advances both the frontier coding performance of GPT-5.2-Codex and the reasoning and... | 400K | \$1.75 | \$14 | Yes |
> | GPT-5.2 Chat | GPT-5.2 Chat (AKA Instant) is the fast, lightweight member of the 5.2 family, optimized for... | 128K | \$1.75 | \$14 | Yes |
> | GPT 5.4 Mini (Fast) | GPT-5.4 Mini brings the strengths of GPT-5.4 to a faster, more efficient model designed for... | 400K | \$1.5 | \$9 | |
> | GPT 5.1 Codex Max | GPT-5.1-Codex-Max is purpose-built for agentic coding. | 400K | \$1.25 | \$10 | Yes |
> | GPT 5.1 Thinking | An upgraded version of GPT-5 that adapts thinking time more precisely to the question to spend... | 400K | \$1.25 | \$10 | |
> | GPT-5 | GPT-5 is OpenAI's flagship language model that excels at complex reasoning, broad real-world... | 400K | \$1.25 | \$10 | Yes |
> | GPT-5-Codex | GPT-5-Codex is a version of GPT-5 optimized for agentic coding tasks in Codex or similar... | 400K | \$1.25 | \$10 | |
> | GPT-5.1 | GPT-5.1 is the latest frontier-grade model in the GPT-5 series, offering stronger general-purpose... | 400K | \$1.25 | \$10 | Yes |
> | GPT-5.1-Codex | GPT-5.1-Codex is a version of GPT-5.1 optimized for agentic coding tasks in Codex or similar... | 400K | \$1.25 | \$10 | Yes |
> | o3 Mini High | OpenAI o3-mini-high is the same model as o3-mini with reasoning\_effort set to high. o3-mini is a... | 200K | \$1.1 | \$4.4 | |
> | o3-mini | o3-mini is OpenAI's most recent small reasoning model, providing high intelligence at the same... | 200K | \$1.1 | \$4.4 | |
> | o4 Mini High | OpenAI o4-mini-high is the same model as o4-mini with reasoning\_effort set to high. OpenAI o4-mini... | 200K | \$1.1 | \$4.4 | |
> | o4-mini | OpenAI's o4-mini delivers fast, cost-efficient reasoning with exceptional performance for its... | 200K | \$1.1 | \$4.4 | |
> | GPT-3.5 Turbo (older v0613) | GPT-3.5 Turbo is OpenAI's fastest model. It can understand and generate natural language or code... | 4K | \$1 | \$2 | Yes |
> | GPT-6.1 Terra | GPT-6.1 Terra is OpenAI s model for visual reasoning, precise computer use, and natural writing... | 1.1M | \$1 | \$5 | Yes |
> | GPT-6.1 Terra Pro | GPT-6.1 Terra Pro is the same underlying model as GPT-6.1 Terra, served with reasoning.mode set to... | 1.1M | \$1 | \$5 | Yes |
> | GPT 5.4 Mini | GPT-5.4 Mini brings the strengths of GPT-5.4 to a faster, more efficient model designed for... | 400K | \$0.75 | \$4.5 | Yes |
> | GPT-4.1 mini (Fast) | GPT 4.1 mini provides a balance between intelligence, speed, and cost that makes it an attractive... | 1M | \$0.7 | \$2.8 | |
> | GPT Audio Mini | A cost-efficient version of GPT Audio. The new snapshot features an upgraded decoder for more... | 128K | \$0.6 | \$2.4 | |
> | GPT-5 mini (Fast) | GPT-5 mini is a cost optimized model that excels at reasoning/chat tasks. It offers an optimal... | 400K | \$0.45 | \$3.6 | |
> | GPT 5.6 Luna (Fast) | GPT-5.6 Luna is a fast, affordable GPT-5.6 model that brings strong capability at the lowest cost... | 1.1M | \$0.4 | \$2.4 | |
> | GPT-4.1 mini | GPT 4.1 mini provides a balance between intelligence, speed, and cost that makes it an attractive... | 1M | \$0.4 | \$1.6 | Yes |
> | GPT 5.1 Codex Mini | GPT-5.1 Codex mini is a smaller, faster, and cheaper version of GPT-5.1 Codex. | 400K | \$0.25 | \$2 | Yes |
> | GPT-4o mini (Fast) | GPT-4o mini from OpenAI is their most advanced and cost-efficient small model. It is multi-modal... | 128K | \$0.25 | \$1 | |
> | GPT-5 mini | GPT-5 mini is a cost optimized model that excels at reasoning/chat tasks. It offers an optimal... | 400K | \$0.25 | \$2 | Yes |
> | GPT 5.4 Nano | GPT-5.4 Nano is designed for tasks where speed and cost matter most like classification, data... | 400K | \$0.2 | \$1.25 | Yes |
> | GPT 5.6 Luna | GPT-5.6 Luna is a fast, affordable GPT-5.6 model that brings strong capability at the lowest cost... | 272K | \$0.2 | \$1.2 | Yes |
> | GPT-5.6 Luna Pro | GPT-5.6 Luna Pro is the same underlying model as GPT-5.6 Luna, served with reasoning.mode set to... | 1.1M | \$0.2 | \$1.2 | Yes |
> | GPT-6 Luna (Fast) | GPT-6 Luna is OpenAI's efficient reasoning model for focused, high-volume tasks. It accepts text... | 1.1M | \$0.2 | \$1 | |
> | GPT OSS 120B | Extremely capable general-purpose LLM with strong, controllable reasoning capabilities | 131K | \$0.15 | \$0.6 | Yes |
> | GPT OSS Safeguard 120B | GPT OSS Safeguard 120B is OpenAI s open-weight safety model for content moderation and guardrail... | 128K | \$0.15 | \$0.6 | |
> | GPT-4o mini | GPT-4o mini from OpenAI is their most advanced and cost-efficient small model. It is multi-modal... | 128K | \$0.15 | \$0.6 | Yes |
> | GPT-4o-mini (2024-07-18) | GPT-4o mini is OpenAI's newest model after GPT-4 Omni, supporting both text and image inputs with... | 128K | \$0.15 | \$0.6 | |
> | GPT-4.1 nano | GPT-4.1 nano is the fastest, most cost-effective GPT 4.1 model. | 1M | \$0.1 | \$0.4 | Yes |
> | GPT-6 Luna | GPT-6 Luna is OpenAI's efficient reasoning model for focused, high-volume tasks. It accepts text... | 1.1M | \$0.1 | \$0.5 | Yes |
> | GPT-6 Luna Pro | GPT-6 Luna Pro is the same underlying model as GPT-6 Luna, served with reasoning.mode set to pro... | 1.1M | \$0.1 | \$0.5 | Yes |
> | GPT OSS Safeguard 20B | OpenAI's first open weight reasoning model specifically trained for safety classification tasks... | 131K | \$0.075 | \$0.3 | Yes |
> | GPT-5 nano | GPT-5 nano is a high throughput model that excels at simple instruction or classification tasks. | 400K | \$0.05 | \$0.4 | Yes |
> | GPT OSS 20B | A compact, open-weight language model optimized for low-latency and resource-constrained... | 131K | \$0.018 | \$0.09 | Yes |

> **Google (23)**
>
> | Model | Note | Context | In | Out | Trusted |
> | - | - | - | - | - | - |
> | Gemini 3.1 Pro Preview | This model improves upon Gemini 2.5 Pro and is catered towards challenging tasks, especially those... | 1M | \$2 | \$12 | Yes |
> | Gemini 3.1 Pro Preview Custom Tools | Gemini 3.1 Pro Preview Custom Tools is a variant of Gemini 3.1 Pro that improves tool selection... | 1M | \$2 | \$12 | |
> | Gemini 4 Argon | Gemini 4 Argon is Google s multimodal reasoning model, supporting text, image, and file inputs... | 1M | \$2 | \$10 | Yes |
> | Nano Banana Pro (Gemini 3 Pro Image) | Nano Banana Pro is Google s most advanced image-generation and editing model, built on Gemini 3... | 66K | \$2 | \$12 | Yes |
> | Gemini 3.5 Flash | Google's latest model, highly optimized for coding proficiency and parallel agentic execution... | 1M | \$1.5 | \$9 | Yes |
> | Nano Banana 2.1 | Nano Banana 2.1 (Gemini Nano Banana 2.1) is Google's image generation and editing model on the... | 66K | \$1.5 | \$7.5 | |
> | Gemini 2.5 Pro | Gemini 2.5 Pro is our most advanced reasoning Gemini model, capable of solving complex problems... | 1M | \$1.25 | \$10 | |
> | Gemini 2.5 Pro Preview 06-05 | Gemini 2.5 Pro is Google s state-of-the-art AI model designed for advanced reasoning, coding... | 1M | \$1.25 | \$10 | Yes |
> | Gemini 3.8 Flash | Gemini 3.8 Flash is the high-efficiency, cost-effective powerhouse of the Gemini 3 family. It... | 1M | \$0.825 | \$4.12 | Yes |
> | Gemini 3.6 Flash | Gemini 3.6 Flash delivers higher quality across coding, agentic workflows, and web development... | 1M | \$0.75 | \$3.75 | Yes |
> | Gemini 3.7 Flash | Gemini 3.7 Flash is the high-efficiency, cost-effective powerhouse of the Gemini 3 family. It... | 1M | \$0.75 | \$3.75 | Yes |
> | Gemini 3 Flash | Google's most intelligent model built for speed, combining frontier intelligence with superior... | 1M | \$0.5 | \$3 | |
> | Gemini 3 Flash Preview | Gemini 3 Flash Preview is a high speed, high value thinking model designed for agentic workflows... | 1M | \$0.5 | \$3 | Yes |
> | Nano Banana 2 | Gemini 3.1 Flash Image, a.k.a. "Nano Banana 2," is Google s latest state of the art image... | - | \$0.5 | \$60 | |
> | Gemini 2.5 Flash | Gemini 2.5 Flash is a thinking model that offers great, well-rounded capabilities. It is designed... | 1M | \$0.3 | \$2.5 | |
> | Gemini 3.5 Flash Lite | Gemini 3.5 Flash Lite features upgraded agentic capabilities, making the model ideal for subagents... | 1M | \$0.3 | \$2.5 | Yes |
> | Gemini 3.1 Flash Lite | Gemini 3.1 Flash Lite outperforms 2.5 Flash Lite on overall quality and lands close to 2.5 Flash... | 1M | \$0.25 | \$1.5 | Yes |
> | Gemini 3.1 Flash Lite Preview | Gemini 3.1 Flash Lite Preview is Google's high-efficiency model optimized for high-volume use... | 1M | \$0.25 | \$1.5 | |
> | Gemma 4 31B IT | Gemma 4 31B is engineered to tackle the most demanding enterprise workloads and complex reasoning... | 262K | \$0.2 | \$0.4 | Yes |
> | Gemini 2.5 Flash Lite | Gemini 2.5 Flash-Lite is a balanced, low-latency model with configurable thinking budgets and tool... | 1M | \$0.1 | \$0.4 | |
> | Google Gemma 4 26B A4B | Gemma is a family of open models built by Google DeepMind. Gemma 4 models are multimodal, handling... | 262K | \$0.09 | \$0.3 | Yes |
> | Gemma 3 27B | Gemma 3 introduces multimodality, supporting vision-language input and text outputs. It handles... | 131K | \$0.08 | \$0.35 | Yes |
> | Gemma 3 12B | Gemma 3 introduces multimodality, supporting vision-language input and text outputs. It handles... | 131K | \$0.05 | \$0.15 | Yes |

> **xAI (6)**
>
> | Model | Note | Context | In | Out | Trusted |
> | - | - | - | - | - | - |
> | Grok 4.5 | Grok 4.5 is a model from SpaceXAI with frontier performance on coding, knowledge work, and STEM. | 500K | \$2 | \$6 | Yes |
> | Grok 4.6 | Grok 4.6 is a model from SpaceXAI with frontier performance on coding, knowledge work, and STEM... | 500K | \$2 | \$6 | Yes |
> | Grok 4.7 | Grok 4.7 is SpaceXAI's flagship model for coding, agentic tasks, and knowledge work, succeeding... | 500K | \$2 | \$6 | Yes |
> | Grok 4.20 | Grok 4.20 is a reasoning model from SpaceXAI with industry-leading speed and agentic tool calling... | 2M | \$1.25 | \$2.5 | Yes |
> | Grok 4.3 | Grok 4.3 is a reasoning model from SpaceXAI. It accepts text and image inputs with text output... | 1M | \$1.25 | \$2.5 | Yes |
> | Grok Build 0.1 | Grok Build 0.1 is SpaceXAI s fast coding model trained specifically for agentic software... | 256K | \$1 | \$2 | Yes |

> **DeepSeek (19)**
>
> | Model | Note | Context | In | Out | Trusted |
> | - | - | - | - | - | - |
> | DeepSeek-R1 | DeepSeek-R1 provides customers a state-of-the-art reasoning model, optimized for general reasoning... | 64K | \$1.35 | \$5.4 | |
> | DeepSeek V4 Pro | DeepSeek-V4 series incorporate several key upgrades in architecture and optimization: (1) a hybrid... | 1M | \$0.955 | \$1.91 | Yes |
> | DeepSeek V4 Pro 0813 | This is the 8/13 updated weights version of DeepSeek V4 Pro. | 1M | \$0.66 | \$1.98 | Yes |
> | DeepSeek V3.2 Thinking | DeepSeek-V3.2 from DeepSeek harmonizes high computational efficiency with superior reasoning and... | 128K | \$0.62 | \$1.85 | |
> | DeepSeek V4.1 Flash 1017 Fast | Higher-throughput variant of DeepSeek V4.1 Flash 1017 served by Baseten for multimodal reasoning... | 1M | \$0.6 | \$2.4 | |
> | DeepSeek V4.1 Flash Fast | Higher-throughput variant of DeepSeek V4.1 Flash for responsive coding, reasoning, and agentic... | 1M | \$0.6 | \$2.4 | |
> | R1 0528 | May 28th update to the original DeepSeek R1 Performance on par with OpenAI o1, but open-sourced... | 164K | \$0.5 | \$2.15 | Yes |
> | DeepSeek V3 0324 | DeepSeek V3, a 685B-parameter, mixture-of-experts model, is the latest iteration of the flagship... | 164K | \$0.29 | \$1.14 | Yes |
> | DeepSeek V3.2 | DeepSeek-V3.2 from DeepSeek harmonizes high computational efficiency with superior reasoning and... | 164K | \$0.28 | \$0.42 | Yes |
> | DeepSeek V3.1 Terminus | DeepSeek-V3.1 Terminus is an update to DeepSeek V3.1 that maintains the model's original... | 164K | \$0.27 | \$1 | Yes |
> | DeepSeek V3.2 Exp | DeepSeek-V3.2-Exp is an experimental large language model released by DeepSeek as an intermediate... | 164K | \$0.27 | \$0.41 | Yes |
> | DeepSeek V3 | DeepSeek-V3 is the latest model from the DeepSeek team, building upon the instruction following... | 128K | \$0.257 | \$1.03 | Yes |
> | DeepSeek V3.1 | DeepSeek-V3.1 is a large hybrid reasoning model (671B parameters, 37B active) that supports both... | 164K | \$0.25 | \$0.95 | Yes |
> | DeepSeek V3.1 | DeepSeek-V3.1 is post-trained on the top of DeepSeek-V3.1-Base, which is built upon the original... | 164K | \$0.25 | \$0.95 | |
> | DeepSeek V4 Flash Vision Exp | DeepSeek-V4-Flash-Vision-Exp is an experimental multimodal model that combines the agentic... | 1M | \$0.216 | \$0.647 | Yes |
> | DeepSeek V4.1 Flash 1017 | DeepSeek V4.1 Flash 1017 served by Fireworks for multimodal reasoning, coding, tool calling, and... | 1M | \$0.15 | \$0.6 | Yes |
> | DeepSeek V4.1 Flash | DeepSeek V4.1 Flash is an AI model from DeepSeek designed for fast, efficient multimodal... | 1M | \$0.14 | \$0.42 | Yes |
> | DeepSeek V4 Flash 0731 | Reasoning model, trusted for Google-connected agents, 1M+ context. | 1M | \$0.119 | \$0.238 | Yes |
> | DeepSeek V4 Flash | Reasoning model, trusted for Google-connected agents, 1M+ context. | 1M | \$0.091 | \$0.182 | Yes |

> **Moonshot (11)**
>
> | Model | Note | Context | In | Out | Trusted |
> | - | - | - | - | - | - |
> | Kimi K3 Fast | Fast version of Kimi s flagship model for long-horizon coding and end-to-end knowledge work, with... | 1M | \$4.5 | \$22.5 | |
> | Kimi K3.1 Pro Highspeed | Kimi K3.1 Pro Highspeed is the high-speed variant of Moonshot AI s frontier Kimi K3.1 Pro model... | 1M | \$4 | \$20 | |
> | Kimi K3.1 Pro | Kimi K3.1 Pro is a frontier model from Moonshot AI for complex reasoning, coding, and agentic... | 1M | \$2 | \$10 | Yes |
> | Kimi K2.7 Code High Speed | Kimi K2.7 Code HighSpeed is the high-speed version of Kimi K2.7 Code, the same model as Kimi K2.7... | 262K | \$1.9 | \$8 | |
> | Kimi K2.6 | Kimi K2.6 demonstrates particularly strong performance in long-horizon coding tasks and produces... | 262K | \$0.65 | \$3.41 | Yes |
> | Kimi K2 0905 | Kimi K2 0905 is the September update of Kimi K2 0711. It is a large-scale Mixture-of-Experts (MoE)... | 262K | \$0.6 | \$2.5 | Yes |
> | Kimi K2 Thinking | Kimi K2 Thinking is Moonshot AI s most advanced open reasoning model to date, extending the K2... | 262K | \$0.6 | \$2.5 | Yes |
> | Kimi K2 Instruct | Kimi K2 is a state-of-the-art mixture-of-experts (MoE) language model with 32 billion activated... | 131K | \$0.57 | \$2.3 | Yes |
> | Kimi K2.7 Code | Kimi-K2.7-Code is a coding model from Moonshot AI. It has improved coding & agent performance over... | 262K | \$0.56 | \$3.1 | Yes |
> | Kimi K2.5 | kimi-k2.5 is Kimi's most versatile model to date, featuring a native multimodal architecture that... | 262K | \$0.5 | \$2.5 | Yes |
> | Kimi K3 | Kimi s flagship model for long-horizon coding and end-to-end knowledge work, with a 1M-token... | 1M | \$0.179 | \$20 | Yes |

> **Qwen (50)**
>
> | Model | Note | Context | In | Out | Trusted |
> | - | - | - | - | - | - |
> | Qwen 3.8 Max Prime | Qwen 3.8 Max Prime is the high-speed edition of Qwen 3.8 Max, retaining its 2.4-trillion-parameter... | 1M | \$4 | \$12 | |
> | Qwen 3.8 Max | Qwen 3.8 Max is a 2.4-trillion-parameter MoE model delivering a comprehensive leap in coding and... | 1M | \$2 | \$6 | |
> | Qwen3.8 2.4T A95B | Qwen 3.8 Max is a 2.4-trillion-parameter MoE model delivering a comprehensive leap in coding and... | 1M | \$2 | \$6 | Yes |
> | Qwen3.8 Max 0902 | Qwen3.8-Max-0902 is an upgraded snapshot of qwen3.8-max. | 1M | \$2 | \$6 | |
> | Qwen 3.7 Max | Qwen3.7 is a next-generation flagship model designed for the agent-centric era, with its core... | 1M | \$1.48 | \$4.42 | |
> | Qwen 3 Max Thinking | Compared with the snapshot as of September 23, 2025, the Qwen-3 series Max model in this release... | 262K | \$1.2 | \$6 | |
> | Qwen3.5 397B A17B | The Qwen3.5 series 397B-A17B native vision-language model is built on a hybrid architecture that... | 262K | \$0.5 | \$3.6 | Yes |
> | Qwen 3.6 27B | The Qwen3.6 35B-A3B native vision-language model is built on a hybrid architecture that integrates... | 262K | \$0.45 | \$2.7 | Yes |
> | Qwen3 235B A22B Thinking 2507 | Qwen3-235B-A22B-Thinking-2507 is a high-performance, open-weight Mixture-of-Experts (MoE) language... | 131K | \$0.45 | \$3.5 | Yes |
> | Qwen 3.5 Plus | The Qwen3.5 native vision-language series Plus models are built on a hybrid architecture that... | 1M | \$0.4 | \$2.4 | |
> | Qwen2.5 72B Instruct | Qwen2.5 72B is the latest series of Qwen large language models. Qwen2.5 brings the following... | 33K | \$0.36 | \$0.4 | Yes |
> | Qwen 3.6 Plus | The Qwen3.6 native vision-language Plus series models demonstrate exceptional performance on par... | 1M | \$0.325 | \$1.95 | |
> | Qwen 3.7 Plus | Among the Qwen3.7 series, the cost-effective Plus model builds on its text capabilities while... | 1M | \$0.32 | \$1.28 | |
> | Qwen3 Coder 480B A35B Instruct | Mixture-of-experts LLM with advanced coding and reasoning capabilities | 262K | \$0.3 | \$1 | Yes |
> | Qwen3.5 Plus 2026-04-20 | Qwen3.5 Plus (April 2026) is a large-scale multimodal language model from Alibaba. It accepts... | 1M | \$0.3 | \$1.8 | |
> | Qwen3 VL 30B A3B Thinking | Qwen3-VL-30B-A3B-Thinking is a multimodal model that unifies strong text generation with visual... | 131K | \$0.29 | \$1 | Yes |
> | Qwen-Plus | Qwen-Plus, based on the Qwen2.5 foundation model, is a 131K context model with a balanced... | 1M | \$0.26 | \$0.78 | |
> | Qwen3.5 Plus 2026-02-15 | The Qwen3.5 native vision-language series Plus models are built on a hybrid architecture that... | 1M | \$0.26 | \$1.56 | |
> | Qwen3.5-122B-A10B | The Qwen3.5 122B-A10B native vision-language model is built on a hybrid architecture that... | 262K | \$0.26 | \$2.08 | Yes |
> | Qwen3.5-27B | The Qwen3.5 27B native vision-language Dense model incorporates a linear attention mechanism... | 262K | \$0.26 | \$2.6 | Yes |
> | Qwen3 VL 8B Instruct | Qwen3-VL-8B-Instruct is a multimodal vision-language model from the Qwen3-VL series, built for... | 131K | \$0.25 | \$0.75 | Yes |
> | Qwen3 235B A22B | Standard model, 262K context. | 262K | \$0.22 | \$0.88 | |
> | Qwen3 VL 235B A22B Instruct | Qwen3-VL-235B-A22B Instruct is an open-weight multimodal model that unifies strong text generation... | 131K | \$0.21 | \$1.9 | Yes |
> | Qwen3 VL 235B A22B Instruct | The Qwen3 series VL models has been comprehensively upgraded in areas such as visual coding and... | 262K | \$0.2 | \$0.88 | |
> | Qwen3 Coder Flash | Qwen3 Coder Flash is Alibaba's fast and cost efficient version of their proprietary Qwen3 Coder... | 1M | \$0.195 | \$0.975 | |
> | Qwen3.6 Flash | Qwen3.6 Flash is a fast, efficient language model from Alibaba's Qwen 3.6 series. It supports... | 1M | \$0.188 | \$1.12 | |
> | Qwen3.5-9B | Qwen3.5-9B is a multimodal foundation model from the Qwen3.5 family, designed to deliver strong... | 262K | \$0.17 | \$0.25 | Yes |
> | Qwen 3 32B | Qwen3-32B is a world-class model with comparable quality to DeepSeek R1 while outperforming... | 128K | \$0.15 | \$0.6 | |
> | Qwen 3 Coder 30B A3B Instruct | Efficient coding specialist balancing performance with cost-effectiveness for daily development... | 262K | \$0.15 | \$0.6 | |
> | Qwen 3.8 Flash | Qwen3.8-Flash is Qwen s fast, cost-efficient multimodal model, combining advanced reasoning and... | 1M | \$0.15 | \$0.47 | |
> | Qwen 3.8 Omni Flash | Qwen3.8-Omni-Flash is Alibaba s native multimodal model for understanding text, images, audio, and... | 1M | \$0.15 | \$0.47 | |
> | Qwen3 Next 80B A3B Thinking | Qwen3-Next uses a highly sparse MoE design: 80B total parameters, but only \~3B activated per... | 131K | \$0.15 | \$1.2 | Yes |
> | Qwen3 VL 30B A3B Instruct | Qwen3-VL-30B-A3B-Instruct is a multimodal model that unifies strong text generation with visual... | 131K | \$0.15 | \$0.6 | Yes |
> | Qwen3.5-35B-A3B | The Qwen3.5 Series 35B-A3B is a native vision-language model designed with a hybrid architecture... | 262K | \$0.14 | \$1 | Yes |
> | Qwen3 14B | Qwen3-14B is a dense 14.8B parameter causal language model from the Qwen3 series, designed for... | 41K | \$0.12 | \$0.24 | Yes |
> | Qwen3 30B A3B | Qwen3, the latest generation in the Qwen large language model series, features both dense and... | 41K | \$0.12 | \$0.5 | Yes |
> | Qwen3 Coder Next | Qwen3-Coder-Next is an open-weight language model built specifically for coding, with strong... | 262K | \$0.12 | \$0.8 | Yes |
> | Qwen3-14B | Qwen3 is the latest generation of large language models in Qwen series, offering a comprehensive... | 41K | \$0.12 | \$0.24 | |
> | Qwen3-30B-A3B | Qwen3 is the latest generation of large language models in Qwen series, offering a comprehensive... | 41K | \$0.12 | \$0.5 | |
> | Qwen 3.5 Flash | The Qwen3.5 native vision-language Flash models are built on a hybrid architecture that integrates... | 1M | \$0.1 | \$0.4 | |
> | Qwen2.5 7B Instruct | Qwen2.5 7B is the latest series of Qwen large language models. Qwen2.5 brings the following... | 33K | \$0.1 | \$0.2 | Yes |
> | Qwen3 30B A3B Instruct 2507 | Qwen3-30B-A3B-Instruct-2507 is a 30.5B-parameter mixture-of-experts language model from Qwen, with... | 128K | \$0.1 | \$0.3 | Yes |
> | Qwen3.6 35B A3B | Qwen3.6-35B-A3B is an open-weight multimodal model from Alibaba Cloud with 35 billion total... | 262K | \$0.1 | \$0.9 | Yes |
> | Qwen3 235B A22B Instruct 2507 | Qwen3-235B-A22B-Instruct-2507 is a multilingual, instruction-tuned mixture-of-experts language... | 262K | \$0.09 | \$0.55 | Yes |
> | Qwen3 Next 80B A3B Instruct | A new generation of open-source, non-thinking mode model powered by Qwen3. This version... | 262K | \$0.09 | \$1.1 | Yes |
> | Qwen3 32B | Qwen3-32B is a dense 32.8B parameter causal language model from the Qwen3 series, optimized for... | 41K | \$0.08 | \$0.28 | Yes |
> | Qwen3 Coder 30B A3B Instruct | Qwen3-Coder-30B-A3B-Instruct is a 30.5B parameter Mixture-of-Experts (MoE) model with 128 experts... | 262K | \$0.07 | \$0.28 | Yes |
> | Qwen3.5-Flash | The Qwen3.5 native vision-language Flash models are built on a hybrid architecture that integrates... | 1M | \$0.065 | \$0.26 | |
> | Qwen3.8 27B | Built on the architectural foundation of Qwen3.5, Qwen3.8 delivers substantial gains across... | 1M | \$0.039 | \$1.25 | Yes |
> | Qwen 3.7 Flash | The Qwen3.7 native vision-language Flash model series delivers a comprehensive upgrade over... | 1M | \$0.03 | \$0.13 | |

> **Mistral (26)**
>
> | Model | Note | Context | In | Out | Trusted |
> | - | - | - | - | - | - |
> | Mistral Large | This is Mistral AI's flagship model, Mistral Large 2 (version mistral-large-2407). It's a... | 128K | \$2 | \$6 | Yes |
> | Mistral Large 2407 | This is Mistral AI's flagship model, Mistral Large 2 (version mistral-large-2407). It's a... | 131K | \$2 | \$6 | Yes |
> | Mixtral 8x22B Instruct | Mistral's official instruct fine-tuned version of Mixtral 8x22B. It uses 39B active parameters out... | 66K | \$2 | \$6 | Yes |
> | Mistral Medium Latest | Mistral's frontier-class multimodal model optimized for agentic and coding use cases. | 262K | \$1.5 | \$7.5 | Yes |
> | Mistral Large 4 | Mistral Large 4 is Mistral AI's open-weight, natively multimodal model, available in public... | 524K | \$1.36 | \$4.18 | |
> | Mistral Large 4 | Mistral Large 4 is a frontier multimodal (text and image input) model from Mistral AI built for... | 1M | \$1.36 | \$4.18 | Yes |
> | Mistral Large 3 | Mistral Large 3 2512 is Mistral s most capable model to date. It has a sparse mixture-of-experts... | 256K | \$0.5 | \$1.5 | |
> | Mistral Large 3 2512 | Mistral Large 3 2512 is Mistral s most capable model to date, featuring a sparse... | 262K | \$0.5 | \$1.5 | Yes |
> | Devstral 2 2512 | Devstral 2 is a state-of-the-art open-source model by Mistral AI specializing in agentic coding... | 262K | \$0.4 | \$2 | Yes |
> | Mistral Medium 3 | Mistral Medium 3 is a high-performance enterprise-grade language model designed to deliver... | 131K | \$0.4 | \$2 | Yes |
> | Mistral Medium 3.1 | Mistral Medium 3.1 is an updated version of Mistral Medium 3, which is a high-performance... | 131K | \$0.4 | \$2 | Yes |
> | Mistral Small 3.1 24B | Mistral Small 3.1 24B Instruct is an upgraded variant of Mistral Small 3 (2501), featuring 24... | 128K | \$0.351 | \$0.555 | Yes |
> | Codestral 2508 | Mistral's cutting-edge language model for coding released end of July 2025. Codestral specializes... | 256K | \$0.3 | \$0.9 | Yes |
> | Mistral Codestral | Mistral's cutting-edge language model for coding released end of July 2025, Codestral specializes... | 128K | \$0.3 | \$0.9 | |
> | Ministral 14B | Ministral 3 14B is the largest model in the Ministral 3 family, offering state-of-the-art... | 256K | \$0.2 | \$0.2 | |
> | Ministral 3 14B 2512 | The largest model in the Ministral 3 family, Ministral 3 14B offers frontier capabilities and... | 262K | \$0.2 | \$0.2 | Yes |
> | Saba | Mistral Saba is a 24B-parameter language model specifically designed for the Middle East and South... | 33K | \$0.2 | \$0.6 | Yes |
> | Ministral 3 8B 2512 | A balanced model in the Ministral 3 family, Ministral 3 8B is a powerful, efficient tiny language... | 262K | \$0.15 | \$0.15 | Yes |
> | Ministral 8B | A more powerful model with faster, memory-efficient inference, ideal for complex workflows and... | 128K | \$0.15 | \$0.15 | |
> | Mistral Small | Mistral Small currently runs Mistral Small 4, a multimodal model combining instruction following... | 32K | \$0.15 | \$0.6 | |
> | Mistral Small 4 | Mistral Small 4 is the next major release in the Mistral Small family, unifying the capabilities... | 256K | \$0.15 | \$0.6 | Yes |
> | Ministral 3 3B 2512 | The smallest model in the Ministral 3 family, Ministral 3 3B is a powerful, efficient tiny... | 131K | \$0.1 | \$0.1 | Yes |
> | Ministral 3B | A compact, efficient model for on-device tasks like smart assistants and local analytics, offering... | 128K | \$0.1 | \$0.1 | |
> | Voxtral Small 24B 2507 | Voxtral Small is an of Mistral Small 3, incorporating state-of-the-art audio input capabilities... | 32K | \$0.1 | \$0.3 | Yes |
> | Mistral Small 3.2 24B | Mistral-Small-3.2-24B-Instruct-2506 is an updated 24B parameter model from Mistral optimized for... | 256K | \$0.075 | \$0.2 | Yes |
> | Mistral Nemo 12B | 12B model trained jointly by Mistral AI and NVIDIA, it significantly outperforms existing models... | 128K | \$0.029 | \$0.03 | Yes |

> **Meta (17)**
>
> | Model | Note | Context | In | Out | Trusted |
> | - | - | - | - | - | - |
> | Muse Wonder | Muse Wonder is a frontier reasoning model from Meta, designed for complex personal and... | 1M | \$4 | \$20 | |
> | Muse Wonder | Muse Wonder is Meta s model for frontier coding, long-running agents, and advanced computer use... | 1M | \$4 | \$20 | |
> | Muse Spark 1.1 | Muse Spark 1.1 is strongest at agentic performance, tool use, and computer use. It does well on... | 1M | \$1.25 | \$4.25 | |
> | Muse Spark 1.2 | A coding-optimized model purpose-built for agentic workflows. Improvements to code generation... | 1M | \$1.25 | \$4.25 | |
> | Muse Spark 1.3 | Muse Spark 1.3 is Meta s multimodal reasoning model for long-horizon agentic and coding workflows... | 1M | \$1.25 | \$4.25 | |
> | Muse Spark Open | Muse Spark Open is Meta s open-weight reasoning model for coding and agentic workflows. Its sparse... | 1M | \$1.25 | \$4.25 | Yes |
> | Llama 3.1 70B Instruct | An update to Meta Llama 3 70B Instruct that includes an expanded 128K context length... | 128K | \$0.72 | \$0.72 | |
> | Llama 3.3 70B Instruct | Where performance meets efficiency. This model supports high-performance conversational AI... | 128K | \$0.72 | \$0.72 | |
> | Llama 3.1 70B Instruct | Meta's latest class of model (Llama 3.1) launched with a variety of sizes & flavors. This 70B... | 131K | \$0.4 | \$0.4 | Yes |
> | Muse Glimmer 30B | Reasoning model, trusted for Google-connected agents, 131K context. | 131K | \$0.3 | \$1.2 | Yes |
> | Llama 3.1 8B Instruct | An update to Meta Llama 3 8B Instruct that includes an expanded 128K context length... | 128K | \$0.22 | \$0.22 | |
> | Llama 4 Maverick 17B Instruct | As a general purpose LLM, Llama 4 Maverick contains 17 billion active parameters, 128 experts, and... | 1M | \$0.188 | \$0.652 | Yes |
> | Llama 3.3 70B Instruct | The Meta Llama 3.3 multilingual large language model (LLM) is a pretrained and instruction tuned... | 131K | \$0.1 | \$0.32 | Yes |
> | Llama 4 Scout 17B Instruct | Llama 4 Scout is the best multimodal model in the world in its class and is more powerful than our... | 328K | \$0.1 | \$0.3 | Yes |
> | Muse Spark 1.2 Contributor | A coding-optimized model with pricing designed for builders. Same model, same capabilities , up to... | 1M | \$0.1 | \$0.2 | |
> | Muse Spark 1.3 Contributor | Muse Spark 1.3 Contributor is Meta s cost-optimized API option for agentic coding workflows. It... | 1M | \$0.1 | \$0.2 | |
> | Llama 3.1 8B Instruct | Meta's latest class of model (Llama 3.1) launched with a variety of sizes & flavors. This 8B... | 131K | \$0.02 | \$0.04 | Yes |

> **MiniMax (10)**
>
> | Model | Note | Context | In | Out | Trusted |
> | - | - | - | - | - | - |
> | MiniMax M2.5 High Speed | M2.5 highspeed: Same performance, faster and more agile (output speed approximately 100 tps) | 205K | \$0.6 | \$2.4 | |
> | MiniMax M2.7 High Speed | M2.7 Highspeed: Same performance, faster and more agile (output speed approximately 100 tps) | 205K | \$0.6 | \$2.4 | |
> | MiniMax M1 | MiniMax-M1 is a large-scale, open-weight reasoning model designed for extended context and... | 1M | \$0.55 | \$2.2 | Yes |
> | MiniMax M2.1 | MiniMax 2.1 is MiniMax's latest model, optimized specifically for in coding, tool use, instruction... | 205K | \$0.3 | \$1.2 | Yes |
> | MiniMax M2.1 Lightning | MiniMax-M2.1-lightning is a faster version of MiniMax-M2.1, offering the same performance but with... | 205K | \$0.3 | \$2.4 | |
> | MiniMax M2.5 | MiniMax-M2.5 is a SOTA large language model designed for real-world productivity. It is capable of... | 197K | \$0.3 | \$1.2 | Yes |
> | MiniMax M2.7 | M2.7 delivers outstanding performance in real-world software engineering, including end-to-end... | 205K | \$0.3 | \$1.2 | Yes |
> | MiniMax M3 | MiniMax-M3 is a frontier-class foundation model that unites the three capabilities defining... | 524K | \$0.3 | \$1.2 | Yes |
> | MiniMax M2 | MiniMax-M2 redefines efficiency for agents. It is a compact, fast, and cost-effective MoE model... | 205K | \$0.255 | \$1.02 | Yes |
> | MiniMax-M3.1-Flash | MiniMax-M3.1-Flash is MiniMax s multimodal M-series model for agent reasoning, tool use, coding... | 1M | \$0.15 | \$0.6 | Yes |

> **NVIDIA (7)**
>
> | Model | Note | Context | In | Out | Trusted |
> | - | - | - | - | - | - |
> | Nemotron 3 Ultra | A 550B parameter (55B active) open reasoning model from NVIDIA, built for long-running agent... | 512K | \$0.6 | \$2.4 | Yes |
> | NVIDIA Nemotron 3 Super 120B A12B | NVIDIA Nemotron 3 Super is a 120B-parameter open hybrid MoE model, activating just 12B parameters... | 262K | \$0.3 | \$2.4 | Yes |
> | Nvidia Nemotron Nano 12B V2 VL | The model is an auto-regressive vision language model that uses an optimized transformer... | 131K | \$0.2 | \$0.6 | |
> | NVIDIA Nemotron 3.5 Super VL | NVIDIA Nemotron 3.5 Super VL is a vision-language model for reasoning, coding, tool use, and... | 1M | \$0.15 | \$0.5 | Yes |
> | Nemotron 3.5 Lightning 30B | NVIDIA-Nemotron-3.5-Lightning-30B-A3B-NVFP4 is a large language model (LLM) trained by NVIDIA. The... | 262K | \$0.07 | \$0.2 | Yes |
> | Nemotron 3 Nano 30B A3B | NVIDIA Nemotron 3 Nano is an open reasoning model optimized for fast, cost-efficient inference... | 262K | \$0.06 | \$0.24 | Yes |
> | Nvidia Nemotron Nano 9B V2 | NVIDIA-Nemotron-Nano-9B-v2 is a large language model (LLM) trained from scratch by NVIDIA, and... | 131K | \$0.06 | \$0.23 | |

> **Amazon (9)**
>
> | Model | Note | Context | In | Out | Trusted |
> | - | - | - | - | - | - |
> | Nova Premier 1.0 | Amazon Nova Premier is the most capable of Amazon s multimodal models for complex reasoning tasks... | 1M | \$2.5 | \$12.5 | Yes |
> | Nova Pro | A highly capable multimodal model with the best combination of accuracy, speed, and cost for a... | 300K | \$0.8 | \$3.2 | |
> | Nova Pro 1.0 | Amazon Nova Pro 1.0 is a capable multimodal model from Amazon focused on providing a combination... | 300K | \$0.8 | \$3.2 | Yes |
> | Nova 2 Lite | Nova 2 Lite is a fast, cost-effective reasoning model for everyday workloads that can process... | 1M | \$0.33 | \$2.75 | |
> | Nova 2 Lite | Nova 2 Lite is a fast, cost-effective reasoning model for everyday workloads that can process... | 1M | \$0.3 | \$2.5 | Yes |
> | Nova Lite | A very low cost multimodal model that is lightning fast for processing image, video, and text... | 300K | \$0.06 | \$0.24 | |
> | Nova Lite 1.0 | Amazon Nova Lite 1.0 is a very low-cost multimodal model from Amazon that focused on fast... | 300K | \$0.06 | \$0.24 | Yes |
> | Nova Micro | A text-only model that delivers the lowest latency responses at very low cost. | 128K | \$0.035 | \$0.14 | |
> | Nova Micro 1.0 | Amazon Nova Micro 1.0 is a text-only model that delivers the lowest latency responses in the... | 128K | \$0.035 | \$0.14 | Yes |

> **Cohere (4)**
>
> | Model | Note | Context | In | Out | Trusted |
> | - | - | - | - | - | - |
> | Command A | Command A is Cohere's most performant model to date, excelling at tool use, agents, retrieval... | 256K | \$2.5 | \$10 | |
> | Command R+ (08-2024) | command-r-plus-08-2024 is an update of the Command R+ with roughly 50% higher throughput and 25%... | 128K | \$2.5 | \$10 | |
> | Command A+ | Command A+ is Cohere's flagship model for enterprise agentic workflows. It accepts text and image... | 192K | \$0.3 | \$1.5 | Yes |
> | Command R (08-2024) | command-r-08-2024 is an update of the Command R with improved performance for multilingual... | 128K | \$0.15 | \$0.6 | |

> **ByteDance (6)**
>
> | Model | Note | Context | In | Out | Trusted |
> | - | - | - | - | - | - |
> | Seed 2.1 Turbo | Seed 2.1 Turbo is a multimodal model from ByteDance Seed for coding and long-horizon agent... | 262K | \$0.5 | \$2.5 | Yes |
> | Seed-2.0-Code | Seed 2.0 Code is a model from ByteDance Seed optimized for agentic coding. It is suited for... | 262K | \$0.5 | \$3 | Yes |
> | Seed 1.6 | Seed 1.6 is a general-purpose model released by the ByteDance Seed team. It incorporates... | 262K | \$0.25 | \$2 | Yes |
> | Seed-2.0-Lite | Seed-2.0-Lite is a versatile, cost-efficient enterprise workhorse that delivers strong multimodal... | 262K | \$0.25 | \$2 | Yes |
> | Seed-2.0-Mini | Seed-2.0-mini targets latency-sensitive, high-concurrency, and cost-sensitive scenarios... | 262K | \$0.1 | \$0.4 | Yes |
> | Seed 1.6 Flash | Seed 1.6 Flash is an ultra-fast multimodal deep thinking model by ByteDance Seed, supporting both... | 262K | \$0.075 | \$0.3 | Yes |

> **OpenRouter (4)**
>
> | Model | Note | Context | In | Out | Trusted |
> | - | - | - | - | - | - |
> | Auto Router (Beta) | The experimental version of our Auto Router where we test new improvements. Use it to get the... | 2M | - | - | |
> | OpenRouter Auto | The Auto Router automatically selects the best model for your prompt, powered by the wisdom of the... | 2M | - | - | |
> | OpenRouter Fusion | Fusion turns your prompt into a small multi-model deliberation. A panel of expert models (see... | 1M | - | - | |
> | Pareto Code Router | The Pareto Router maintains a tiered shortlist of strong coding models, ranked by Artificial... | 2M | - | - | |

> **Aion-labs (5)**
>
> | Model | Note | Context | In | Out | Trusted |
> | - | - | - | - | - | - |
> | Aion 3.5 | Aion 3.5 is a multi-model roleplaying and storytelling system from AionLabs, built on the GLM... | 262K | \$3 | \$6 | |
> | Aion-3.0 | Aion-3.0 is a multi-model roleplaying and storytelling system from AionLabs, built on the GLM... | 131K | \$3 | \$6 | |
> | Aion-2.0 | Aion-2.0 is a variant of DeepSeek V3.2 optimized for immersive roleplaying and storytelling. It is... | 131K | \$0.8 | \$1.6 | |
> | Aion 3.5 Mini | Aion 3.5 Mini is a multi-model roleplaying and storytelling system from AionLabs, built on the GLM... | 262K | \$0.7 | \$1.4 | |
> | Aion-3.0-Mini | Aion-3.0 Mini is a multi-model roleplaying and storytelling system from AionLabs, built on the... | 131K | \$0.7 | \$1.4 | |

> **Arcee-ai (1)**
>
> | Model | Note | Context | In | Out | Trusted |
> | - | - | - | - | - | - |
> | Trinity Large Thinking | Trinity-Large-Thinking is a reasoning-optimized variant of Arcee AI's Trinity-Large family , a... | 262K | \$0.25 | \$0.8 | |

> **Bytedance (3)**
>
> | Model | Note | Context | In | Out | Trusted |
> | - | - | - | - | - | - |
> | Seed 2.1 Turbo | Seed 2.1 Turbo is ByteDance s multimodal language model supporting text, image, and video inputs... | 262K | \$0.5 | \$2.5 | |
> | Bytedance Seed 1.8 | Bytedance Seed 1.8 features stronger multimodal understanding and agent capabilities. The model... | 256K | \$0.25 | \$2 | |
> | Seed 1.6 | ByteDance's new multimodal deep-thinking model, supporting both text and visual inputs with... | 256K | \$0.25 | \$2 | |

> **Callstack (1)**
>
> | Model | Note | Context | In | Out | Trusted |
> | - | - | - | - | - | - |
> | Apex | Apex is Callstack s agentic coding model for mobile and web development, optimized for React... | 262K | \$0.5 | \$3 | |

> **Elorian (1)**
>
> | Model | Note | Context | In | Out | Trusted |
> | - | - | - | - | - | - |
> | Glimpse | Glimpse is a multimodal visual reasoning model from Elorian, built for visual understanding and... | 262K | \$0.6 | \$3 | Yes |

> **Fireworks (1)**
>
> | Model | Note | Context | In | Out | Trusted |
> | - | - | - | - | - | - |
> | Ember-1 | Ember-1 is Fireworks research-preview reasoning model built on Kimi K3 for coding and agentic... | 1M | \$3 | \$15 | Yes |

> **Ibm-granite (1)**
>
> | Model | Note | Context | In | Out | Trusted |
> | - | - | - | - | - | - |
> | Granite 4.2 8B | Granite 4.2 8B is a dense reasoning model from IBM. It is suited for mathematics, code generation... | 131K | \$0.06 | \$0.25 | Yes |

> **Inception (4)**
>
> | Model | Note | Context | In | Out | Trusted |
> | - | - | - | - | - | - |
> | Mercury 2 | A diffusion-based reasoning LLM that generates text via parallel refinement (not token-by-token)... | 128K | \$0.25 | \$0.75 | Yes |
> | Mercury Coder Small Beta | Mercury Coder Small is ideal for code generation, debugging, and refactoring tasks with minimal... | 32K | \$0.25 | \$1 | |
> | Mercury 3 | Mercury 3 is Inception s most capable model for agentic workflows and real-time applications. It... | 1M | \$0.1 | \$0.35 | Yes |
> | Mercury 2.5 | Mercury 2.5 is Inception s diffusion-based reasoning model for chat, agents, and structured... | 260K | \$0.04 | \$0.15 | Yes |

> **Inclusionai (5)**
>
> | Model | Note | Context | In | Out | Trusted |
> | - | - | - | - | - | - |
> | Ling 3.0 Flash Fin | Ling 3.0 Flash Fin is InclusionAI s finance- MoE language model, combining 124 billion total... | 256K | \$0.075 | \$0.22 | |
> | Ling 3.0 Flash Sante | Ling-3.0-Flash-Sante is inclusionAI s language model specialized for health and medicine, built on... | 256K | \$0.075 | \$0.22 | |
> | Ling 3.1 Flash | Ling 3.1 Flash is InclusionAI's hybrid reasoning language model for coding, multi-step analysis... | 1M | \$0.075 | \$0.225 | Yes |
> | Ling 3.0 Flash VL | Ling 3.0 Flash VL builds on Ling 3.0 Flash with stronger language capabilities, native visual... | 131K | \$0.06 | \$0.18 | Yes |
> | Ling 3.0 Flash | Ling 3.0 Flash is designed with token efficiency and production-scale agentic inference as key... | 262K | \$0.021 | \$0.063 | Yes |

> **Interfaze (1)**
>
> | Model | Note | Context | In | Out | Trusted |
> | - | - | - | - | - | - |
> | Interfaze Beta | Interfaze is an AI model built on a new architecture that merges specialized DNN/CNN models with... | 1M | \$1.5 | \$3.5 | |

> **Liquid (3)**
>
> | Model | Note | Context | In | Out | Trusted |
> | - | - | - | - | - | - |
> | LFM2.5-2.6B | LFM2.5-2.6B is a compact reasoning model from Liquid AI. It is suited for agent workflows, data... | 128K | \$0.35 | \$0.5 | |
> | LFM2.5-24B-A2B | LFM2.5-24B-A2B is a mixture-of-experts reasoning model from Liquid AI with 24B total and 2.45B... | 128K | \$0.05 | \$0.7 | Yes |
> | LFM2.5-24B-A2B | LFM2.5-24B-A2B is a text-generation model from Liquid AI with a 128,000-token context window. | 128K | \$0.05 | \$0.7 | |

> **Meituan (2)**
>
> | Model | Note | Context | In | Out | Trusted |
> | - | - | - | - | - | - |
> | LongCat 2.0 | LongCat 2.0 is a sparse mixture-of-experts language model from Meituan, with 48B active parameters... | 1M | \$0.3 | \$1.2 | |
> | LongCat 2.5 Preview | LongCat-2.5-Preview is Meituan s multimodal reasoning model for coding and agentic workflows, with... | 1M | \$0.3 | \$1.2 | |

> **Mixedbread (1)**
>
> | Model | Note | Context | In | Out | Trusted |
> | - | - | - | - | - | - |
> | Toast 1 | Built for knowledge-intensive questions, multi-step retrieval, and evidence synthesis. Toast 1 can... | 131K | \$0.3 | \$0.72 | |

> **Nex-agi (1)**
>
> | Model | Note | Context | In | Out | Trusted |
> | - | - | - | - | - | - |
> | Nex-N2.5-Pro | Nex-N2.5 is an agentic model built to turn goals into working, verified outcomes. Its core... | 262K | \$0.075 | \$0.25 | |

> **Perceptron (1)**
>
> | Model | Note | Context | In | Out | Trusted |
> | - | - | - | - | - | - |
> | Perceptron Mk1.5 | Perceptron Mk1.5 is Perceptron's embodied reasoning model for physical agents. It accepts text... | 37K | \$0.15 | \$1.5 | Yes |

> **Poolside (2)**
>
> | Model | Note | Context | In | Out | Trusted |
> | - | - | - | - | - | - |
> | Laguna S 2.1 | Laguna S 2.1 is Poolside's new open-weight model for agentic coding and long-horizon work. | 1M | \$0.09 | \$0.18 | |
> | Laguna XS 2.1 | Laguna XS 2.1 is the latest coding agent model in the 33B-A3B category from Poolside and a step... | 262K | \$0.06 | \$0.12 | |

> **Prism-ml (2)**
>
> | Model | Note | Context | In | Out | Trusted |
> | - | - | - | - | - | - |
> | Ternary Bonsai 2 9B | Ternary Bonsai 2 9B is PrismML s compact, open-weight language model for conversational AI... | 262K | \$0.08 | \$0.13 | |
> | Ternary Bonsai 2 27B | Bonsai 2 27B is a 27B-parameter reasoning model from PrismML derived from Qwen3.8-27B. It supports... | 262K | \$0.075 | \$0.5 | |

> **Quiverai (2)**
>
> | Model | Note | Context | In | Out | Trusted |
> | - | - | - | - | - | - |
> | Arrow 2 Telos | Higher-refinement Arrow 2 variant combining Arrow's speed with frontier-model reasoning. | 131K | \$6 | \$30 | |
> | Arrow 2 | SVG generation model with agentic refinement and tool calling. | 131K | \$4 | \$20 | |

> **Reflection (1)**
>
> | Model | Note | Context | In | Out | Trusted |
> | - | - | - | - | - | - |
> | Beam 1.0 501B A23B | Beam 1.0 is Reflection's open-weight reasoning model for coding and long-horizon agentic... | 1M | \$0.25 | \$1 | Yes |

> **Rekaai (1)**
>
> | Model | Note | Context | In | Out | Trusted |
> | - | - | - | - | - | - |
> | Reka Edge | Reka Edge is an extremely efficient 7B multimodal vision-language model that accepts... | 16K | \$0.1 | \$0.1 | Yes |

> **Relace (1)**
>
> | Model | Note | Context | In | Out | Trusted |
> | - | - | - | - | - | - |
> | Relace Search | The relace-search model uses 4-12 view\_file and grep tools in parallel to explore a codebase and... | 256K | \$1 | \$3 | Yes |

> **Sakana (5)**
>
> | Model | Note | Context | In | Out | Trusted |
> | - | - | - | - | - | - |
> | Fugu Ultra | Fugu Ultra coordinates a deeper pool of expert agents to maximize answer quality on hard... | 1M | \$5 | \$30 | |
> | Fugu Ultra v2 | Fugu Ultra coordinates a deeper pool of expert agents to maximize answer quality on hard... | 1M | \$5 | \$30 | |
> | Fugu Max | Fugu Max orchestrates our largest pool of models to push the cost-performance Pareto frontier. It... | 1M | \$2 | \$6 | |
> | Sakana Namazu | Sakana Namazu is a Japanese-specialized reasoning model from Sakana AI, based on Kimi K2.6 with... | 262K | \$0.95 | \$4 | |
> | Sakana Namazu | Sakana Namazu is a Japanese-specialized LLM that combines a deep understanding of Japanese culture... | 256K | \$0.95 | \$4 | |

> **Spacexai (11)**
>
> | Model | Note | Context | In | Out | Trusted |
> | - | - | - | - | - | - |
> | Grok 4.5 | SpaceXAI's smartest model with frontier performance on coding, knowledge work, and STEM. | 500K | \$2 | \$6 | |
> | Grok 4.6 | Grok 4.6 builds on Grok 4.5 with a particular focus on long-running agents and more ambitious... | 500K | \$2 | \$6 | |
> | Grok 4.7 | Grok 4.7 is SpaceXAI s advanced AI model for coding and professional knowledge work, built to... | 500K | \$2 | \$6 | |
> | Grok 4.20 Beta Non-Reasoning | Grok 4.20 Beta is the newest flagship model from xAI with industry-leading speed and agentic tool... | 2M | \$1.25 | \$2.5 | |
> | Grok 4.20 Beta Reasoning | Grok 4.20 Beta is the newest flagship model from xAI with industry-leading speed and agentic tool... | 2M | \$1.25 | \$2.5 | |
> | Grok 4.20 Multi Agent Beta | Multiple agents collaborate in parallel to perform deep research tasks. | 2M | \$1.25 | \$2.5 | |
> | Grok 4.20 Multi-Agent | Multiple agents collaborate in parallel to perform deep research tasks. | 2M | \$1.25 | \$2.5 | |
> | Grok 4.20 Non-Reasoning | Grok 4.20 Beta is the newest flagship model from xAI with industry-leading speed and agentic tool... | 2M | \$1.25 | \$2.5 | |
> | Grok 4.20 Reasoning | Grok 4.20 Beta is the newest flagship model from xAI with industry-leading speed and agentic tool... | 2M | \$1.25 | \$2.5 | |
> | Grok 4.3 | Grok 4.3 is a new model matching the scale of Grok 4.20 with an improved architecture and a... | 1M | \$1.25 | \$2.5 | |
> | Grok Build 0.1 | xAI's fast coding model trained specifically for agentic coding. | 256K | \$1 | \$2 | |

> **Stealth (1)**
>
> | Model | Note | Context | In | Out | Trusted |
> | - | - | - | - | - | - |
> | Glyph Cluster | Glyph Cluster is an anonymous reasoning model in early access. Prompts and outputs will be... | 256K | - | - | |

> **Stepfun (3)**
>
> | Model | Note | Context | In | Out | Trusted |
> | - | - | - | - | - | - |
> | Step 5 Preview | Step 5 Preview is StepFun s flagship AI model for agentic coding, professional knowledge work, and... | 1M | \$1 | \$2.7 | Yes |
> | Step 3.7 Flash | StepFun s flagship multimodal reasoning model. Powered by a 198B-parameter / 11B-activation sparse... | 256K | \$0.2 | \$1.15 | Yes |
> | Step 3.5 Flash | Step 3.5 Flash is StepFun's most capable open-source foundation model. Built on a sparse Mixture... | 262K | \$0.1 | \$0.3 | Yes |

> **Tencent (3)**
>
> | Model | Note | Context | In | Out | Trusted |
> | - | - | - | - | - | - |
> | Tencent Hy4 Preview | Tencent Hy4 preview is Tencent Hy s open-source large language model, featuring 770B total... | 1M | \$0.834 | \$2.5 | Yes |
> | Hy3 preview | Hy3 preview is a high-efficiency Mixture-of-Experts model from Tencent designed for agentic... | 262K | \$0.18 | \$0.6 | |
> | Hy3 | Built for real-world business scenarios, Hy3 features a 295B/21B active MoE architecture, native... | 262K | \$0.132 | \$0.528 | Yes |

> **Thinkingmachines (2)**
>
> | Model | Note | Context | In | Out | Trusted |
> | - | - | - | - | - | - |
> | Inkling | Inkling is a multimodal MoE model (975B total, 41B active, 256k context) reasoning over text... | 524K | \$1 | \$4.05 | Yes |
> | Inkling Small | Inkling-Small is a lighter-weight model with 12B active parameters, trained with a similar recipe... | 524K | \$0.45 | \$1.2 | Yes |

> **Typesafe (1)**
>
> | Model | Note | Context | In | Out | Trusted |
> | - | - | - | - | - | - |
> | Jev Router | Jev Router picks the best model and reasoning effort for each request, balancing quality, speed... | 1M | - | - | |

> **Unbiased (2)**
>
> | Model | Note | Context | In | Out | Trusted |
> | - | - | - | - | - | - |
> | Pareto 26.9 | Pareto is a multimodal composite model built for research, coding, and agentic workflows, while... | 262K | \$2.5 | \$7.5 | |
> | Pareto 26.10 | Pareto is a multimodal composite model built for research, coding, and agentic workflows, while... | 1M | \$0.8 | \$3.2 | Yes |

> **Upstage (3)**
>
> | Model | Note | Context | In | Out | Trusted |
> | - | - | - | - | - | - |
> | Solar Pro 3 | Solar Pro 3 is Upstage's powerful Mixture-of-Experts (MoE) language model. With 102B total... | 131K | \$0.15 | \$0.6 | |
> | Solar Pro 4 | Solar Pro 4 is Upstage's cost-efficient large language model, featuring a 524K context window. It... | 524K | \$0.15 | \$0.6 | Yes |
> | Solar Mini 4 | Solar Mini 4 is Upstage's compact, cost-efficient language model, a 35B-parameter... | 524K | \$0.05 | \$0.2 | Yes |

> **Xiaomi (5)**
>
> | Model | Note | Context | In | Out | Trusted |
> | - | - | - | - | - | - |
> | MiMo V2.6 Pro UltraSpeed | MiMo V2.6 Pro UltraSpeed is Xiaomi's accelerated inference offering for MiMo V2.6 Pro, designed... | 1M | \$4.35 | \$8.7 | |
> | MiMo V2.5 Pro | MiMo V2.5 Pro delivers significant improvements over its predecessor, MiMo-V2-Pro, in general... | 1M | \$0.48 | \$1.8 | Yes |
> | MiMo V2.6 Pro | MiMo V2.6 Pro is Xiaomi's flagship multimodal reasoning model for complex software engineering... | 1.1M | \$0.435 | \$0.87 | Yes |
> | MiMo M2.5 | A native full-modal model supporting text, image, video, and audio understanding, with powerful... | 1M | \$0.4 | \$2 | Yes |
> | MiMo V2.6 Flash | MiMo V2.6 Flash is Xiaomi's efficient multimodal reasoning model for coding, automation, and... | 1.1M | \$0.14 | \$0.28 | Yes |

> **Zhipu (19)**
>
> | Model | Note | Context | In | Out | Trusted |
> | - | - | - | - | - | - |
> | GLM 5.2 Fast | GLM-5.2-Fast-Preview is the high-speed variant of Zhipu AI's GLM-5.2, with 1M context and... | 1M | \$2.8 | \$8.8 | |
> | GLM 5.3 Prime | GLM-5.3-Prime is the high-speed variant of Z.ai's GLM-5.3, inheriting its full capabilities while... | 1M | \$2.8 | \$8.8 | |
> | GLM 5.3 Fast | Speed-optimized version of Z.AI s GLM-5.3 agentic coding model, built for responsive, real-time... | 1M | \$2.1 | \$6.6 | |
> | GLM 5.1 | GLM-5.1 delivers a major leap in coding capability, with particularly significant gains in... | 203K | \$1.4 | \$4.4 | Yes |
> | GLM 5 Turbo | GLM 5 Turbo is a foundation model deeply optimized for the OpenClaw scenario. It has been... | 200K | \$1.2 | \$4 | Yes |
> | GLM 5V Turbo | GLM-5V-Turbo is Z.AI s first multimodal coding foundation model, built for vision-based coding... | 203K | \$1.2 | \$4 | Yes |
> | GLM 5 | GLM 5 is a frontier-class, general-purpose large language model optimized for complex systems... | 205K | \$1 | \$3.2 | Yes |
> | GLM 4.5 | GLM-4.5 and GLM-4.5-Air are our latest flagship models, purpose-built as foundational models for... | 131K | \$0.6 | \$2.2 | Yes |
> | GLM 4.5V | Built on the GLM-4.5-Air base model, GLM-4.5V inherits proven techniques from GLM-4.1V-Thinking... | 66K | \$0.6 | \$1.8 | Yes |
> | GLM 4.6 | As the latest iteration in the GLM series, GLM-4.6 achieves comprehensive enhancements across... | 203K | \$0.55 | \$2.2 | Yes |
> | GLM 4.7 | GLM-4.7 is a powerful next-generation coding model that delivers major gains in multilingual... | 205K | \$0.54 | \$1.98 | Yes |
> | GLM 5.3 FlashX | GLM-5.3-FlashX is the high-speed serving option for Z.ai s native multimodal coding model... | 1M | \$0.37 | \$1.25 | Yes |
> | GLM 4.6V | GLM-4.6V is a large multimodal model designed for high-fidelity visual understanding and... | 131K | \$0.3 | \$0.9 | Yes |
> | GLM 5.3 Flash | GLM-5.3-Flash is Z.ai s native multimodal coding model, featuring 320B total parameters, 18B... | 1M | \$0.15 | \$0.5 | Yes |
> | GLM 4.5 Air | GLM-4.5 and GLM-4.5-Air are our latest flagship models, purpose-built as foundational models for... | 131K | \$0.13 | \$0.85 | Yes |
> | GLM 5.3 | GLM 5.3 delivers comprehensive advancements in complex software engineering and agent... | 1M | \$0.07 | \$1.4 | Yes |
> | GLM 4.7 Flash | GLM-4.7-Flash balances high performance with efficiency, making it the perfect lightweight... | 203K | \$0.061 | \$0.4 | Yes |
> | GLM 4.7 FlashX | GLM-4.7-Flash balances high performance with efficiency, making it the perfect lightweight... | 200K | \$0.06 | \$0.4 | |
> | GLM 5.2 | GLM-5.2 delivers powerful coding capabilities, usable 1M-context support, and continued strengths... | 1M | \$0.014 | \$7 | Yes |

## Related

> **[Models](/docs/account/models)**
>
> How to pick, switch, and keep spend in check.

> **[Trusted models](/docs/agents/trusted-models)**
>
> Why Google connections lock an agent to a smaller set.

> **[Pricing](/docs/account/pricing)**
>
> Credit balance and at-cost model usage.

> **[Usage](/docs/account/usage)**
>
> See which models are drawing down credit.


# Billing
Source: https://getskydive.com/docs/account/pricing

Find the right plan for you or your team

Skydive is subscription-based. Paid plans put **usage credit** on a **dollar balance**: the plan fee for the period equals the credit you get (1:1). Agents spend down that balance as they work. When it runs low, you can buy more, turn on auto top-up, or move up a tier.

You manage this in the app under [**Billing**](https://www.skydive.com/settings/billing), with a **Monthly / Yearly** toggle on the plan cards.

> **Model usage at cost**
>
> **We do not mark up model usage.** What model providers charge is what draws your **credit balance**. No hidden premium on model rates.
>
> Your plan fee still buys the product (agents, seats, platform). Model spend passes through at cost against your credit balance.

## Plans

What you see when choosing a plan:

> **Free**
>
> **\$0** / month, pay as you go
>
> Where every workspace starts, with no card required.
>
> * Add credit anytime, agents resume
> * Pay as you go from your balance
> * Auto-refills when you run low
> * Set a monthly cap anytime

> **Starter**
>
> **\$20** / month
>
> For individuals getting started.
>
> * Unlimited agents
> * Unlimited seats
> * More usage included (\$20 credit / month)
> * Pay as you go after
> * Email support

> **Team**
>
> **\$200** / month
>
> Everything in Starter, plus more room for a group.
>
> * Unlimited agents
> * Unlimited seats
> * More usage included (\$200 credit / month)
> * Pay as you go after
> * Priority support

> **Enterprise**
>
> **Custom**
>
> Everything in Team, plus:
>
> * SSO and SAML
> * Advanced security
> * Audit logs
> * Custom contracts
> * Dedicated support

> [!TIP]
>
> Starter and Team prices are **flat for the whole workspace**, not per seat. All plans, including Free, have unlimited seats.

## Compare plans

| | Free | Starter | Team | Enterprise |
| - | :-: | :-: | :-: | :-: |
| **Price** | \$0 / mo, pay as you go | \$20 / mo | \$200 / mo | Custom |
| **Included credit** | What you buy | \$20 / mo | \$200 / mo | Custom |
| **Agents** | Yes | Unlimited | Unlimited | Unlimited |
| **Seats** | Unlimited | Unlimited | Unlimited | Unlimited |
| **Pay as you go** | ✓ | ✓ | ✓ | ✓ |
| **SSO and SAML** | ✕ | ✕ | ✕ | ✓ |
| **Support** | Community | Email | Priority | Dedicated |

Every workspace starts on Free with a small amount of credit so you can build your first agent. After that you buy credit, subscribe to a plan, or both.

## Monthly or yearly

On the billing page, switch **Monthly** or **Yearly**.

* **Starter yearly:** \$240 / year (12 × \$20), with **\$264** of usage credit for the year (10% bonus credit).
* **Team yearly:** \$2,400 / year (12 × \$200), with **\$2,640** of usage credit for the year (same 10% bonus).

Yearly plans charge and grant credit up front for the full year. The fee is 12 × monthly, with no extra discount on the fee. The incentive is the bonus credit.

## How credits work

Your agents spend from a credit balance. Usage comes from:

> **Model usage (at cost)**
>
> What your agents spend on AI models as they reason and respond. Billed **at cost** with **no markup**: provider rates pass through to your **credit balance**. More capable [models](/docs/account/models) and longer tasks use more.

> **Compute and services**
>
> Sandbox time (browsing, running code, hosting apps) plus metered services your agents call, such as search and email.

On Starter and Team, the subscription fee **is** the included credit for that period (1:1). Starter at \$20/month puts \$20 of credit on the balance each month. Team at \$200/month puts \$200.

That credit covers real usage. **LLM tokens are charged at cost**, with no upcharge on model rates. Compute and other metered services also draw the balance. The no-markup promise is specifically for model token spend.

When included credit runs low:

* **Buy usage credit** as a one-time purchase (presets start at \$50).
* **Auto-refill** when the balance runs low (you control the threshold, amount, and optional monthly cap).
* Or **upgrade** to a higher plan.

The [Usage dashboard](/docs/account/usage) shows spend over time, per agent, and against your balance.

## Controlling spend

Owners and admins set this from two places on [Usage & billing](https://www.skydive.com/settings/billing): **Billing** for how the workspace refills, and the **Spend controls** tab for monthly budgets and who can change them.

### Auto-refill

On the Billing tab, set the balance **threshold** that triggers a refill and the **restore balance** it fills back to. The minimum threshold is \$5. The restore balance must be at least \$10 above it.

Paid plans also carry a small buffer so a run mid-flight isn't cut off the instant credit hits zero: **\$5 on Starter** and **\$10 on Team**. Pay-as-you-go workspaces have no cushion. Once the buffer is used, new work waits for more balance.

### Spend controls

The **Spend controls** tab is for owners and admins. It sets a monthly workspace budget, a default cap for every agent, extra caps for specific agents, members, or teams, and who on the team can see or change those numbers.

* **Monthly budget.** Optional workspace ceiling for the billing period. When spend crosses it, new agent runs pause until you raise or remove the budget, or the period resets. Runs already in flight are not cut off mid-task. Minimum \$1, maximum \$1,000,000.
* **Warn admins at.** Optional heads-up email before the budget is reached (50%, 75%, 80%, or 90%). If you leave this off, the first email is the one that says the budget ran out.
* **Default agent limit.** Optional monthly cap that applies to every agent unless you set a different one below. Agents with no extra cap spend against the workspace budget.
* **Additional limits.** Caps on top of the workspace budget for individual agents, members, teams, or people within a team. A row that says **No additional limit** is not uncapped. It still spends against the workspace budget. You can pick a preset or type a custom dollar amount.
* **What members can do.** Choose who can see their own spend (everyone, admins, or owners only) and who can change an agent's limit (anyone who can edit the agent, admins, or owners only).

Limits reset when the billing period resets, not necessarily on the 1st of the month. Members who are allowed to see their own spend get a limit and a running total on their Usage page.

Full walkthrough on [Spend Control](/docs/account/cost-tips#spend-controls).

## When billing pauses your work

If a workspace can't cover its usage, agents drop to a **free floor**: a small amount of activity keeps running so your agents aren't fully dark, with tight limits (a few runs per hour and per day, capped tokens, the smallest compute tier, and a monthly compute ceiling). Members see a notice to update billing, and full capacity returns the moment there is balance again. The free floor covers your agents' public apps too: an agent's webserver and inbound webhooks keep serving at the floor's compute tier instead of showing a billing-paused page. An owner or admin can fix it. What to do depends on why it paused:

| Why it paused | What resumes it |
| - | - |
| Out of credit (pay as you go) | Buy credit or subscribe to a plan. |
| Out of credit (paid) | Add credit, or let auto-refill run. |
| Payment method declined | Update the card on the billing page. |
| Hard spend limit reached | Raise or remove the limit, or wait for the next period. |

A declined card is retried for a couple of weeks. If it still hasn't gone through after about three weeks, the workspace drops to **pay as you go**. Your data stays; the plan's included credit and paid features stop. Update the card any time to restore the plan.

## Enterprise

Enterprise is custom. Self-serve setup and **Contact sales** are both options on the billing page when you need SSO, advanced security, audit logs, or a paper contract.

[Talk to us](mailto:hello@skydive.com) about Enterprise.

## Promotional credit

If you have a promo code, you can apply it directly from your billing page.

1. Go to [**Settings, Usage & Billing**](https://www.skydive.com/settings/billing).
2. In the **Credit** section, click **Redeem** (next to the Buy credit button).
3. Enter your code and submit.

Promo credit is added to your balance immediately and shows as a separate line in the Credit section. One code per person.

## Add-ons

Some capabilities bill on top of the plan, per agent:

| Add-on | Price |
| - | - |
| **iMessage dedicated line** | \$100 / month per agent |

A shared iMessage pool line is free and available to every workspace. The dedicated add-on is for agents that need their own exclusive number.

Add-ons use your workspace card, do not grant usage credit, and show as their own invoice line.

## Frequently asked

> **Is Team priced per seat?**
>
> No. Team is \$200 per month for the whole workspace, with unlimited seats. All plans have unlimited seats. Seats do not multiply the price.

> **Can I just buy credit without a subscription?**
>
> Yes. **Buy usage credit** on the billing page adds balance so agents can keep working, with pay as you go and optional auto-refill after that. Subscribing to Starter or Team adds a recurring included credit.

> **What happens when I run out of credit?**
>
> Agents don't go fully dark. They drop to a **free floor**: a small amount of activity keeps running under tight limits (a few runs per hour and per day, capped tokens, the smallest compute tier, and a monthly compute ceiling) so an agent can still respond while you sort out billing. Full capacity returns as soon as there is balance again. Buy credit, let auto-refill run, or upgrade. You will see it coming on the Usage dashboard and the credit remaining indicator on Billing.

> **Can I change plans later?**
>
> Yes. Moving to a higher monthly amount usually takes effect right away (often prorated for the rest of the period). Moving down, or off a yearly plan, generally waits until the end of the current period so prepaid credit is not cut short.

> **Do yearly plans cost less per month?**
>
> The dollar total is 12 × the monthly price. Yearly plans grant **10% extra usage credit** for the year as the incentive.

> **Do you mark up LLM usage?**
>
> No. **Model usage is at cost.** We do not add a markup on LLM tokens. Provider rates pass through against your credit balance. The plan fee is for Skydive (agents, seats, platform). Token spend is separate and not padded.

> **[Track your usage](/docs/account/usage)**
>
> See what your agents are consuming against your balance.

> **[Talk about Enterprise](mailto:hello@skydive.com)**
>
> Custom contracts, security review, and a dedicated contact.


# Profile & settings
Source: https://getskydive.com/docs/account/profile

Your account and how agents recognize you across channels

Profile & settings is your personal account on Skydive: your details, how you sign in, and how agents recognize you across channels. This is separate from an [agent's identity](/docs/agents/persona), which is the agent's own name and avatar.

## Account

Your personal details:

* **Full name**: how you're shown to teammates.
* **Email**: the address tied to your account.
* **Phone number**: the verified number that channel connects key to. See below.
* **Role**: your role in the workspace.
* **Avatar**: the image that represents you.

### Phone number

The **Phone number** section shows the number on file and whether it's **Verified**, **Unverified**, or **Not set**. This is the number every iMessage connect reserves its line for, so keeping it current matters: a stale number (say, one saved during signup that you no longer text from) silently breaks new iMessage connects.

Click **Add number** or **Change** to update it. Skydive texts a 5-digit code to the new number, and the number is only attached once you enter that code, so a typo can't replace a working number. On success, any connect flow that depends on it picks up the new number right away.

## Linked sign-in methods

The accounts you can use to sign in to Skydive: currently **Google** and **Slack**. Link one to enable one-click sign-in, and manage which are connected here. New accounts sign up with Google or Slack. If you already sign in with an email and password, that keeps working. This controls how *you* log in. It is separate from the services you connect for an *agent* to use.

## Slack

Connect your Skydive account to a Slack workspace so the platform recognizes you there. Install Skydive to your workspace so your agents are reachable in Slack. Once linked, a DM or @-mention from you in Slack ties the conversation to your account.

See [Slack](/docs/channels/slack) for the full guide to using an agent in Slack.

## Appearance

Pick how Skydive looks from the appearance control in your profile menu: **System**, **Light**, or **Dark**. System follows your device setting and switches with it. Your choice is remembered on this browser. The app re-themes the moment you pick.

## Reset onboarding

Under **Advanced** on your account page, **Reset onboarding** sends you back through the welcome setup to create a new agent, the same first-run flow you saw when you joined. Use it to start fresh or walk a teammate through the setup. Your existing agents and workspace stay as they are.

## Deleting your account

**Delete account** is in the **Sign-in & account** section of your account page. It removes you from every workspace and permanently deletes your personal data. Type `delete my account` to confirm. There is no way to undo it.

Before you do:

* A workspace where you're the **only** member is deleted along with you. In a workspace with other members, your agents are reassigned and your membership is removed.
* If you're the **sole owner** of a workspace that still has other members, make someone else an owner in that workspace's member settings first, then retry. Skydive tells you which workspaces need a new owner.
* Your access ends immediately. Most of your data is permanently deleted within 30 days.

To delete an entire workspace (not just your own account), see [deleting a workspace](/docs/workspaces/overview#deleting-a-workspace).


# Usage
Source: https://getskydive.com/docs/account/usage

Track what your agents consume and review your invoice history

The Usage & billing page shows what your agents consume against your credit balance. Use it to see where spend goes, catch surprises early, and decide when to top up or upgrade. Admins can also pull up invoice history here.

## What you can see

> **Usage over time**
>
> Daily charts of consumption, so you can spot trends and spikes.

> **Per-agent breakdown**
>
> Which agents are doing the most work, and costing the most.

> **Against your balance**
>
> How much credit you've used and what's left.

> **By period**
>
> Filter to the time range you care about.

## What "usage" reflects

Paid plans include usage credit equal to the plan fee for the period. Agents draw that **credit balance** down as they work. Spend comes from:

* **Model usage (at cost)**: the AI model tokens your agents use as they think and respond. Charged **at cost**, with **no markup** on model rates. More capable [models](/docs/account/models) and longer tasks use more.
* **Compute and services**: sandbox time (browsing, running code, hosting apps) plus metered services such as search and email.

Model tokens are at cost, so choose the model that fits the job. You are not paying a platform premium on every token. Simpler tasks, sensible model choices, and focused agents all stretch your credit further.

## Staying ahead of limits

When you approach the end of your included credit, top up or move up a tier on [Billing](https://www.skydive.com/settings/billing). Buy usage credit, turn on auto-refill when your plan supports it, or upgrade. See [Pricing](/docs/account/pricing).

Owners and admins set monthly budgets on the **Spend controls** tab. Full walkthrough on [Spend Control](/docs/account/cost-tips).

If you do run out, agents don't go fully dark. They drop to a **free floor**: a small amount of activity keeps running under tight limits (a few runs per hour and per day, capped tokens, the smallest compute tier, and a monthly compute ceiling) so an agent can still respond while you restore billing. While the floor is active, the usage-limits block on this page shows one bar per enforced limit, the 5-hour window and the daily cap, each with its own reset time. Full capacity returns as soon as there is balance again.

While you're on the free floor, the Your plan card on [Billing](https://www.skydive.com/settings/billing) shows a usage limits row: what percent of those limits your agents have used, and either the next reset time or "Limit reached".

> [!TIP]
>
> Use the per-agent breakdown to find a heavy agent. A single misconfigured or overly chatty agent often drives a surprising share of usage. A small tweak to its scope or model reins it in.

## Invoice history

Workspace owners and admins can view a record of past charges on the **Billing** tab of [Usage & billing](https://www.skydive.com/settings/billing). Invoice history sits right below your payment method, in the same page.

Each invoice shows the date, amount, and status (paid, pending, or failed). For Stripe-collected invoices you can download a PDF copy directly from the list.

Invoices reflect completed billing periods only. Your current period's accruing usage shows in the usage chart above, not here.

> **[Compare plans](/docs/account/pricing)**
>
> See what each tier includes and pick the right one.


# Creating agents
Source: https://getskydive.com/docs/agents/creating-agents

Create, configure, and manage your agents

You can run as many agents as your plan allows. Most teams keep a handful of focused agents, one per job, rather than one generalist.

## Create an agent

On your dashboard, click **New agent** and give it:

* **Name**: how you address it and how it signs its work.
* **Purpose**: a sentence or two on what it is for. This seeds its [persona](/docs/agents/persona).

A purpose reads like a job description:

```text theme={null}
Name:    Support Triage
Purpose: Watch our support inbox, draft replies to common questions,
         and escalate anything urgent or unhappy to a human.
```

On confirmation, Skydive provisions a [sandbox](/docs/agents/overview#the-sandbox) with a fresh memory store and the agent's baseline tools. The agent is ready in seconds.

> [!TIP]
>
> Name and scope agents narrowly: "Support Triage," "Weekly Reporting," "Outreach Research." A focused agent is easier to trust, easier to schedule, and builds more useful memory than a catch-all.

## Configure it

Each agent has settings you can adjust anytime:

> **[Identity & persona](/docs/agents/persona)**
>
> Set its name, handle, and avatar, and how it thinks, writes, and decides.

> **[Models](/docs/account/models)**
>
> Choose which AI models it's allowed to use.

> **[Connections](/docs/integrations/overview)**
>
> Authorize the services it needs.

> **[Channels](/docs/channels/overview)**
>
> Add it to Slack, give it an inbox, or a phone line.

> **[Permissions & visibility](/docs/agents/permissions)**
>
> Set who can see and message it.

> **[Schedule](/docs/agents/routines)**
>
> Set recurring or future-triggered runs.

> **[Compute](#compute)**
>
> Size its sandbox.

> **[Agent repository](/docs/agents/repo)**
>
> The git repository backing the agent. Clone it to inspect memory, MCP config, tools, and the persona file.

## Compute

**Compute** sets how much CPU and memory an agent's [sandbox](/docs/agents/overview#the-sandbox) gets. Choose a tier when you create the agent, or change it anytime from its settings.

| Tier | Memory | Good for |
| - | - | - |
| Lite | 2 GB | Light, low-traffic tasks |
| Standard | 8 GB | Balanced, everyday agent work |
| Pro | 16 GB | Code and data-heavy workloads |

Memory is the headline number. Each tier pairs it with a matching amount of CPU.

New agents start at **4 GB** by default. That puts them between Lite and Standard: enough for most everyday tasks without the overhead of a larger machine. Upgrade to Standard or higher from the agent's settings when the job needs more. Tiers above a free ceiling may require a paid plan with billing enabled.

A bigger sandbox does more at once. It also draws down your [compute](/docs/account/usage) usage faster for the same running time. Pick the smallest tier that keeps the agent's work smooth.

Changing an agent's compute rebuilds its sandbox at the new size. The change takes effect on the agent's next run, not mid-task.

## Create a dedicated agent for a big job

Some work deserves its own agent. An agent is a peer with its own sandbox, memory, tools, and channels, not a sub-agent of another. Create a separate agent when a substantial, ongoing responsibility needs a clean identity and its own memory.

An agent can create a peer this way on its own, and archive it when the work is done for good. It can also undo a mistaken archive. It can only touch agents you have edit access to, and it never archives itself.

## Review and history

Every agent keeps a record of its work:

* The [feed](/docs/workspaces/feed) shows actions across all conversations.
* Conversation history lets you revisit any past thread.

## Duplicate an agent

To create a copy of an existing agent, open the **···** menu on its profile and choose **Duplicate**. The copy starts with the same persona, connections, and routines as the original. It has no conversation history and an empty memory, so it's a clean slate with the same configuration.

Duplicating is useful when you want to experiment with a different persona or scope without touching a working agent, or when you need two agents doing the same job in different contexts.

## Retire an agent

When an agent has done its job and you don't need it anymore, choose **Retire** from the **···** menu.

Retiring an agent:

* Stops all its scheduled routines immediately
* Hides it from your active agent list
* Keeps its memory, conversations, and history intact

A retired agent can be restored anytime. Open it from your archived agents list and choose **Reactivate**. Skydive provisions a fresh sandbox from the agent's stored configuration and it's back.

## Reset an agent

Reset returns an agent to the state it was in when it was first created. Use it when you want to start fresh without losing the agent's identity.

**What gets cleared:**

* All conversations and conversation history
* All memory files
* Posts and feed activity
* Scheduled routines
* Files in the shared workspace
* All connected integrations and secrets
* Any customizations made to the agent's repo (reverted to the initial commit)

**What stays:**

* The agent's name, avatar, and handle
* Its place in the workspace

> [!WARNING]
>
> Reset cannot be undone. Everything listed above is permanently deleted.

To reset, open the **···** menu on the agent's profile and choose **Reset agent**.

> **[Version history](/docs/agents/version-history)**
>
> Browse and restore past states of the agent's repo without a full reset.

> **[Give your agent a persona](/docs/agents/persona)**
>
> The single highest-impact thing you can configure.


# Memory
Source: https://getskydive.com/docs/agents/memory

How agents remember across conversations and teammates

Every agent has durable, file-based memory that persists across conversations. Teach it something once and it stays taught.

![Vera arranging memory and skill cards](https://mintcdn.com/skydive/2VR7bKqFZZqW4hyR/images/agents/memory-skills-vera.png?fit=max\&auto=format\&n=2VR7bKqFZZqW4hyR\&q=85\&s=d2a933e6d8fd22987793fc02bc153f19)

## Agent-level, not per-person

Memory is **shared across every conversation** an agent has. When one teammate teaches it something, everyone who talks to it next gets that knowledge. The agent organizes what it knows by subject: a person, a project, a piece of general feedback, an external system.

> [!NOTE]
>
> Memory is shared knowledge, like a team wiki the agent maintains for itself. Because anyone with access to the agent benefits from it, treat it as a shared space, not somewhere to store sensitive personal disclosures or secrets.

## What agents remember

* **About people**: your preferences, role, working style, and recurring needs.
* **About projects**: context, decisions, status, and the things that matter to a body of work.
* **General feedback**: lessons and corrections not tied to one person.
* **Reference**: how external systems, dashboards, and accounts are set up.

## How memories form

You don't manage memory by hand. The agent writes to it as it works. When you state a durable preference, correct it, or establish a fact it will need again, it remembers:

```text theme={null}
Going forward, our fiscal quarters start in February, not January.
```

The next time anyone asks about quarters, the agent already knows.

## Recalling memory

At the start of each conversation, the agent has an index of what it knows. It pulls in the relevant pieces as needed. You can also ask it directly:

```text theme={null}
What do you remember about the Q3 launch?
```

## Correcting or forgetting

Memory is editable. Tell the agent to update or drop something and it will:

```text theme={null}
Forget the old office address. We moved. The new one is 500 Howard St.
```

## The nightly review

Once a day, in a quiet window, your agent runs a reflection pass on its own. It reads back recent conversations, looks for repeated corrections and patterns, and folds durable lessons into its memory. The point is consolidation: merge new lessons into rules it already keeps, tighten anything that drifted, and skip the day entirely when nothing worth keeping surfaced.

This runs by default. To turn it off for one agent, open the agent's profile, choose Dreaming in the menu, and flip the toggle off. Other agents keep theirs unless you change them too.

If your agent has a repo, you can also add a `dreaming.md` file next to `soul.md` to replace the reflection instructions with your own. Leave it out and the agent uses the default guide.

## Across channels

Memory follows the agent, not the channel. Teach it something in Slack and it remembers in web chat and email too. It is one agent with one memory, wherever you reach it.

> [!NOTE]
>
> Each conversation runs in its own sandbox with its own snapshot of memory taken when it started. A fact taught in another conversation becomes visible to an in-progress one once it refreshes, which the agent can do on demand when it matters.

> **[Set deliberate behavior with a persona](/docs/agents/persona)**
>
> Memory is what the agent learns; the persona is how you tell it to behave.

> **[Explore memory files directly](/docs/agents/repo)**
>
> Memory lives in the agent's git repository. Clone it to read, edit, or delete what it knows.


# Overview
Source: https://getskydive.com/docs/agents/overview

The anatomy of a Skydive agent

An agent is the core unit of Skydive. Each one is an independent worker with everything it needs to act on your behalf. Here's what every agent is made of.

## Anatomy of an agent

> **[Identity & persona](/docs/agents/persona)**
>
> Name, handle, and avatar, plus the standing instructions that shape how it thinks, writes, and decides.

> **Sandbox**
>
> A dedicated, isolated cloud machine where the agent runs code, browses, edits files, and hosts apps.

> **[Memory](/docs/agents/memory)**
>
> Durable, file-based knowledge that persists across every conversation and teammate.

> **[Tools](/docs/capabilities/tools-and-mcp)**
>
> Built-in abilities (browser, search, files, web server) plus anything you connect via MCP.

> **[Connections](/docs/integrations/overview)**
>
> Authorized accounts and API keys the agent uses through a secure proxy.

> **[Channels](/docs/channels/overview)**
>
> The channels you reach it on: web, Slack, email, text.

> **[Schedule](/docs/agents/routines)**
>
> Optional recurring or future-triggered runs that fire without a human in the loop.

> **[Agent repository](/docs/agents/repo)**
>
> A private git repository backing the agent. Where its persona, memory, tools, and MCP config live.

## The sandbox

Every agent gets its own cloud sandbox: a Linux machine provisioned on demand and isolated from every other agent. Inside it, the agent can:

* Run shell commands and execute code
* Read, write, and edit files
* Drive a real web browser
* Search and fetch web content
* Host a live web server you can visit
* Install packages and tools it needs

The sandbox is **ephemeral**: it spins up when there is work and shuts down when idle. The agent saves anything it needs to keep (memory, shared files, code) to durable storage automatically. Nothing important is lost between runs. The agent's configuration (its persona, MCP connections, tools, and memory files) lives in its [git repository](/docs/agents/repo).

You can size the sandbox per agent, from a small default up to 16 vCPU for heavy jobs. See [Compute](/docs/agents/creating-agents#compute).

## How it works

**1. You send a message**

On any channel. The agent receives it along with its memory, your identity, and the conversation history.

**2. It plans and acts**

The agent decides what to do, then uses its tools (searching, browsing, writing code, calling APIs) inside its sandbox.

**3. It keeps you in the loop**

Progress streams back as it works. For multi-step jobs it keeps a visible task list. It pauses to ask when it needs a decision or a credential.

**4. It reports back**

You get the result: a message, a file, or a link to a live app. It saves anything worth remembering to memory.

## Managing agents

Your dashboard is where you manage agents. Create them, rename them, adjust their persona, review their [feed](/docs/workspaces/feed) and history, manage their connected services, and set who can reach them. Run as many as your plan allows. Most people run a few specialized agents rather than one that does everything.

Each agent has its own access settings that control who can see, message, and change it, down to a single person or team. See [Permissions & visibility](/docs/agents/permissions).

> **[Create your first agent](/docs/get-started/quickstart)**
>
> Walk through it step by step.


# Permissions & visibility
Source: https://getskydive.com/docs/agents/permissions

Who can see an agent, who can use it, and who can change it

Every agent carries its own access settings. They decide who can find it, who can put it to work, and who can change how it behaves. Open an agent's **⋮** menu and choose **Access** to see and change them.

## What access covers

Access comes in four roles, weakest to strongest: View, Use, Edit, Owner. Each role includes everything to its left.

| | View | Use | Edit | Owner |
| - | :-: | :-: | :-: | :-: |
| Find and open the agent | ✓ | ✓ | ✓ | ✓ |
| Message the agent and put it to work | | ✓ | ✓ | ✓ |
| Change persona, connections, and skills | | | ✓ | ✓ |
| Manage access settings and delete the agent | | | | ✓ |

The agent's creator starts as its owner. Someone can use an agent every day without being able to change how it behaves. That is the normal arrangement for a shared team agent.

## General access

Every agent has a general-access level that applies to everyone in your workspace, and a separate one for people outside it.

| Level | Workspace members | People outside |
| - | - | - |
| **No access** | Can't find the agent | Can't reach it |
| **View** | Can see it | Can't reach it |
| **Use** | Can message it | Can message it |
| **Edit** | Can change it | Can message it |

Use this to answer the broad question: should the whole workspace reach this agent, or nobody outside a short list? General access applies to everyone in the workspace by default. Fine-tune it with grants.

## Grants to people and teams

When general access is too broad, grant a role to specific people or to [teams](/docs/workspaces/teams). A grant gives one person (or every member of a team) the role you pick, on this one agent. Grants stack with general access: whoever has the highest role wins.

Some examples:

* A hiring agent used by recruiting only: no workspace access, then grants for the recruiting team.
* A shared support agent the whole workspace can message, but only two people can edit: general access **Use**, plus **Edit** grants for those two.
* An agent still being built: no access for the workspace, and **View** or **Use** grants for the people reviewing it.

> [!NOTE]
>
> Workspace owners and admins can always see, message, and edit every agent in the workspace, whatever its access settings.

## People outside your workspace

Opening an agent beyond your workspace lets people without a Skydive account email or message it from their own channels. What they can do depends on the general-access level for people outside:

* **Use** (the default): anyone who reaches it can message the agent.
* Any other level: outsiders reach nothing.

> [!WARNING]
>
> A publicly reachable agent can be messaged by anyone who finds it, including unidentified senders over email or text. Use it for agents meant to face customers or the public, not for anything that touches private workspace data.

## Managing access

Open the agent, click the **⋮** menu, and choose **Access**. The dialog lists general access and every person and team grant. Add someone with **Share**, pick their role, and changes apply immediately. Removing someone's access stops them on their next view refresh. Access changes also send the affected person an email.

Only the agent's owner (or a workspace owner or admin) can open the Access dialog.

## Sharing an agent

In the same **⋮** menu, choose **Share** for a link to the agent's profile, plus a profile card you can download as an image. Whether a teammate can open the link follows their access: no access and the link shows nothing; **View** or above and it opens the agent.

## Access and shared memory

Access and [memory](/docs/agents/memory) go together. Everyone who can message an agent draws on the same memory. A widely shared agent is a shared space: knowledge one person teaches it helps everyone. Keep private or sensitive details out of a shared agent, and keep the workspace level at **View** when the work itself shouldn't be visible to the rest of the workspace.

## What an agent itself can reach

Access governs people. What the *agent* can reach is a separate question, answered by the credentials you connect and the proxy that holds them.

> **[Integration permissions](/docs/integrations/permissions)**
>
> How credentials are scoped, and why the agent never sees a raw token.

> **[Roles in your workspace](/docs/workspaces/rbac)**
>
> Owner, admin, and member, and what each can do.

> **[Teams](/docs/workspaces/teams)**
>
> Grouping people, and granting them agent access together.


# Identity & persona
Source: https://getskydive.com/docs/agents/persona

Who the agent is, and how it thinks

Identity is how people recognize the agent. Persona is how it behaves. Both live on this page.

## Identity

Every agent has a name, a handle, an avatar, and a profile. These show up in every conversation, on every channel, and on every piece of work it ships. Change them anytime from the agent's settings.

| Field | What it is |
| - | - |
| Name | How you address it and how it signs its work. |
| Handle | A unique `@handle` across Skydive. Used in mentions, URLs, and the agent's email address (`handle@skydivemail.ai`). |
| Avatar | The image next to its messages. |
| Profile | A public page at `skydive.com/agent/<id>` with its bio, connected channels, and recent work. |

The handle is unique. The name does not have to be.

The avatar is a generated preset by default. From the agent's settings you can upload a photo instead (PNG, JPEG, or WEBP, up to 5 MB), and remove it anytime to fall back to the preset.

A new agent gets a name and a purpose. The purpose seeds its persona. The rest of the identity you set from settings.

## Persona

The persona is the standing instruction that shapes how the agent thinks, writes, and decides. It is not a prompt you paste into a conversation. It is the file the agent reads before every turn.

A tight persona is a job description: who this agent is, who it works for, what good looks like, and what it should never do.

> [!TIP]
>
> Write the persona as if you were briefing a new teammate on their first day. Specific beats exhaustive. A page of real examples beats three pages of principles.

## What to put in it

These are the sections that actually change behavior:

* **Role**: one or two sentences on what this agent is for. "You are the on-call engineer for the payments API," not "You are a helpful assistant."
* **Audience**: who it is talking to, and how. An internal ops agent writes differently than one that emails customers.
* **Voice**: tone, length, words to use, words to avoid. Paste examples of writing you like. The agent will match them.
* **Priorities**: what to optimize for when tradeoffs appear. Speed vs. thoroughness, asking vs. deciding, drafting vs. sending.
* **Guardrails**: hard rules. Never send email without asking. Never merge to main. Never invent a number you did not look up.
* **Context**: the facts it should never have to re-ask. Company name, product, timezone, the people it works with.

Skip anything the agent can learn from doing the work. Memory will pick that up. The persona should stay true even if memory were empty.

## Where it lives

The persona lives in the agent's [repository](/docs/agents/repo) as `soul.md`. Edit it from the agent's settings, or by opening a PR against the repo. Either way, the next conversation picks it up.

Treat it like production code. Small, reviewed changes. If the agent starts behaving oddly, the persona is the first place to look.

## Keep it current

A persona that never changes falls behind the job. When you correct the agent, ask whether the correction belongs in the persona (it should always do this) or in [memory](/docs/agents/memory) (it happened to need this today).

The test: would a new conversation, with no prior context, still do the right thing? If not, update the persona.


# Repo
Source: https://getskydive.com/docs/agents/repo

The git repository that backs every agent. Where memories, tools, MCP config, and persona live.

Every Skydive agent is backed by a private git repository. It's the source of truth for everything that makes the agent *this* agent rather than a generic one: its persona, what it knows, the tools it has, and the code it runs.

You can explore it, edit it, and commit to it just like any other repository.

## Where the repository lives

Inside the agent's sandbox, the repository is mounted at `/home/user/.self`. The agent commits and pushes its changes automatically, so everything it learns and configures is persisted across sandbox restarts.

## What's in it

```
.self/
├── soul.md             # Persona and standing instructions
├── mcp.config.json     # MCP server connections
├── tools/              # Custom in-process tools
├── skills/             # Reusable workflows (one folder per skill)
│   └── <skill>/
│       └── SKILL.md
├── .memory/            # Persistent memory files
│   ├── users/
│   ├── projects/
│   └── reference/
├── Dockerfile          # Baked-in packages and dependencies
├── start.sh            # Credentialed setup that runs at boot
├── webserver.sh        # Starts the agent's web server (if any)
└── index.ts            # Agent harness entrypoint
```

### `soul.md`

The agent's persona. It's the same content you can edit through the dashboard, but here you can version it, diff it, and PR it. The agent reads this at the start of every conversation.

### `mcp.config.json`

Which MCP servers the agent has connected. Add a server here and the agent picks it up on its next turn:

```json theme={null}
{
  "servers": {
    "my-server": {
      "transport": "stdio",
      "command": "npx",
      "args": ["-y", "some-mcp-package"]
    }
  }
}
```

Optional per-server fields pick which tools register:

* `tools`: only these tool names register.
* `excludeTools`: these tool names never register.

If you set both, `tools` applies first, then `excludeTools` removes from what's left. An empty `tools` array is the same as leaving it out. Edits take effect on the next turn; the agent doesn't reconnect the server.

```json theme={null}
{
  "servers": {
    "my-server": {
      "transport": "stdio",
      "command": "npx",
      "args": ["-y", "some-mcp-package"],
      "tools": ["search", "create_issue"],
      "excludeTools": ["delete_all"]
    }
  }
}
```

See [Tools & MCP](/docs/capabilities/tools-and-mcp) for the full connection patterns.

### `tools/`

Custom in-process tools the agent can call. Each `.ts` or `.mjs` file here exports a `ToolDefinition`. The agent picks up new files without a restart. Files starting with `_` or `.` are skipped.

### `skills/`

One subdirectory per skill. Each contains a `SKILL.md` with YAML frontmatter (`name`, `description`) and the workflow instructions. The agent re-scans this on every conversation, so editing a skill file takes effect immediately. See [Skills](/docs/agents/skills).

### `.memory/`

The agent's persistent memory, organized as plain markdown files. The directory structure reflects how the agent groups what it knows:

```
.memory/
├── users/<name>/       # What the agent knows about specific people
├── projects/<name>/    # Context, decisions, and status per project
└── reference/          # How external systems are set up
```

You can read these to understand what the agent has learned, and you can edit or delete them to correct stale facts. The agent reads from this at the start of each conversation. See [Memory](/docs/agents/memory).

### `Dockerfile`

Packages and system dependencies baked into the agent's image. Anything installed here survives sandbox restarts. Use it for tools the agent needs reliably:

```dockerfile theme={null}
RUN apt-get update && apt-get install -y ffmpeg
RUN npm install -g some-global-tool
```

Changes take effect on the next image build.

### `start.sh`

Setup that runs at boot, after credentials are available. This is where you put steps that need secrets or network identity, like cloning a private external repository. The Dockerfile runs at build time with no credentials; `start.sh` runs at runtime with full access to secrets and the git credential helper.

## Exploring an agent's state

To see what an agent remembers, inspect `.memory/` in the repository. To see which tools it has, look in `tools/` and `mcp.config.json`. The persona is in `soul.md`.

```bash theme={null}
# Read the agent's persona
cat .self/soul.md

# See which MCP servers are wired up
cat .self/mcp.config.json

# Browse memory files
ls .self/.memory/
cat .self/.memory/projects/my-project/state.md
```

## Making changes

Changes committed and pushed to the agent's repository take effect on its next run. The agent's sandbox is rebuilt from the Dockerfile on push, so package changes need a push to persist. In-conversation changes (memory writes, tool additions, MCP config edits) are committed by the agent automatically.

> [!WARNING]
>
> The `.memory/` directory uses plain markdown files. Editing them directly is safe, but the agent may overwrite a file if it writes a newer version in its next conversation. Treat direct edits as corrections, not long-term overrides.

## Exporting an agent

An agent's repo is the agent. Persona, memory, skills, tools, and MCP config all live in it, so cloning it is a full export of the agent's setup. Nothing else is needed to reconstruct the agent elsewhere.

In the app, open the agent's **Computer** panel, switch to the **Version History** tab, and click **Copy clone URL**. It mints a git token and copies the full clone URL to your clipboard:

```bash theme={null}
git clone <paste the URL here>
```

The same URL is available from **Manage version history** in the agent's **···** menu. Editors get a read/write token; viewers get a read-only one. See [Version history](/docs/agents/version-history#cloning-the-repo) for the token details.

> [!NOTE]
>
> The repo holds the agent's setup and knowledge. Conversations and workspace data are separate from it.

> **[Tools & MCP](/docs/capabilities/tools-and-mcp)**
>
> How the agent extends its toolset.

> **[Memory](/docs/agents/memory)**
>
> How agents accumulate and recall knowledge.

> **[Skills](/docs/agents/skills)**
>
> Reusable workflows stored in the repository.

> **[Persona](/docs/agents/persona)**
>
> The standing instructions in soul.md.

> **[Platform CLI](/docs/api-reference/platform-cli)**
>
> The runtime commands the agent runs inside its sandbox.


# Routines
Source: https://getskydive.com/docs/agents/routines

Recurring and future-triggered work, no human in the loop

Agents can run on their own schedule. Use routines for anything that should happen **later** or **repeatedly**.

![Penny setting a recurring routine](https://mintcdn.com/skydive/2VR7bKqFZZqW4hyR/images/agents/scheduled-runs-penny.png?fit=max\&auto=format\&n=2VR7bKqFZZqW4hyR\&q=85\&s=8fd3d66392a319fb36a348e911ec4e21)

## What you can schedule

> **Recurring work**
>
> "Every weekday at 9am, summarize what changed in our shared inbox."

> **A one-time future task**
>
> "Remind me at 5pm to send the invoice."

> **Periodic checks**
>
> "Every Monday, check our competitors' pricing pages and flag changes."

> **Reports & digests**
>
> "Send a weekly summary of closed deals every Friday afternoon."

## How it works

Set up a schedule by asking the agent in plain language:

```text theme={null}
Every weekday at 9am Eastern, post a summary of yesterday's support
tickets to our #cs channel.
```

The agent creates the schedule. When it fires, the run happens **with no human in the loop**. The agent acts on the instruction you gave and delivers the result to the channel you specified.

> [!TIP]
>
> Write the instruction with all the context the future run needs. No one is there to clarify. "Summarize yesterday's tickets and post to #cs" beats "do the ticket thing."

## Cadence and timing

* **Recurring**: on a schedule like "every weekday at 9am" or "the first of each month." Time zones are respected: "9am" means 9am where you are.
* **One-time**: a single future moment, like "tomorrow at noon" or a specific date.

## Suggested routines

You don't have to think up every schedule yourself. On the **Routines** page, your agent proposes schedules it could run for you, based on its role, its past work, and its connected services. Each suggestion is a real, tool-using task: check, post, summarize.

Open a suggestion to see what it would do and when. Click **Set it up** to turn it into a live routine, or dismiss it. Suggestions refresh on their own.

## Managing schedules

List, pause, resume, update, or cancel an agent's schedules at any time. Ask in plain language:

```text theme={null}
Pause the daily ticket summary until next week.
```

Each routine has a detail view with its full run history. Every past run links to the conversation it ran in. Open any firing to see exactly what the agent did.

## Choosing a model per routine

By default, each routine runs on the same model the agent uses. You can set a different model on the routine's detail page: open the routine, find the **Model** field, and choose from the picker. That model applies to every firing of that routine only and does not change the agent's general model.

To revert to the agent's default, clear the model on the routine's detail page. You can also set or clear a model from the platform CLI:

```bash theme={null}
platform automation update <id> --model claude-sonnet-5
platform automation update <id> --clear-model
```

## When a routine reads outside data

A routine can read an external service (an inbox, a shared order, an API) and then act on what it read: cancel something, delete something, or post "nothing happened today". Routines run with no one watching, so an empty result needs care before it triggers an action.

An empty read is a claim, not a fact. Every reader has silent failure modes: the wrong endpoint, a view scoped to one account, truncated pages, rate limiting, a stale cache, the wrong time window. None of them error. They all look like "empty".

If a routine can cancel, delete, close, or clear something based on what it read, follow these rules:

1. **Calibrate the reader first.** A reader that has never returned data can't tell "empty" from "broken". Check it once against a case you know has data.
2. **Know whose data you're looking at.** A view of your own slice of shared state isn't a view of the whole. If other people add to what your agent reads, find what they write and check the agent can see it.
3. **Size the blast radius.** If the read is wrong, what does the action destroy, and whose work is it? If the answer is "other people's", empty is not permission.
4. **Treat unknown as hold.** On an empty result your agent can't explain: take no destructive step, leave the thing open, and have the agent flag it to you. A false "nothing today" destroys work. A false hold costs one message.
5. **Cross-check one independent source** before any destructive call: a second endpoint, the thread replies, an admin view. If they disagree, hold.
6. **Default to the reversible branch.** Routines fire with no human present, so the failure mode you build in should never be the one that deletes other people's work.

## Where results go

A routine can deliver to any channel: a Slack channel, an email, or a direct message. There is no live conversation when it fires, so name the destination when you set up the schedule.

> **[Pair it with a skill](/docs/agents/skills)**
>
> Schedule a codified workflow to run a whole process automatically.


# Skills
Source: https://getskydive.com/docs/agents/skills

Reusable workflows your agent can follow

A **skill** is a reusable, codified workflow: instructions, and optionally scripts and templates, that an agent follows whenever a matching task comes up. Define the process once instead of explaining it every time.

## When to use a skill

Use a skill for a **repeatable procedure** with steps worth getting right consistently:

* "How we format and publish the weekly report"
* "Our process for triaging an inbound bug"
* "The steps to onboard a new customer in our tools"

If you have explained the same multi-step process more than once, make it a skill.

## What a skill contains

> **Instructions**
>
> The step-by-step workflow the agent follows.

> **Scripts**
>
> Optional helper scripts the workflow runs.

> **Templates & references**
>
> Boilerplate, examples, and detailed docs the steps point to.

## Skills vs. persona vs. memory

| | Skill | [Persona](/docs/agents/persona) | [Memory](/docs/agents/memory) |
| - | - | - | - |
| **Answers** | How to do a specific task | How to behave generally | What it has learned |
| **Triggered by** | A matching task | Always on | Relevant context |
| **Shape** | A documented procedure | Standing instructions | Accumulated facts |

A skill is the playbook for one job; the persona is the agent's character; memory is its accumulated knowledge.

## Built-in skills

Every agent ships with four skills pre-installed.

**webhook-receiver** covers the full setup handshake for receiving inbound HTTP events from external services: fast acknowledgment, HMAC signature verification, durable cross-run deduplication, and the two-party provider-console flow (endpoint registration, signing secret, validation ping). It works with any provider, including Stripe, GitHub, Linear, and Intercom. Ask an agent to build a webhook receiver and it loads the skill automatically.

**link-wallet** gives your agents access to human-approved online purchases via Stripe Link. The wallet belongs to you, the owner. You log in once and approve every single spend from your Link app, so no card number ever lands in the sandbox. The agent uses `link-cli auth login` to surface a URL and verification phrase, you approve in the Link app, and the CLI returns a one-time virtual card for the purchase. Spend limits are set by Link ($500 per transaction, $500 per day). Use `link-cli --test` during development to verify the flow without a real charge.

**gmail** loads when an agent works through a connected Gmail inbox at any real volume. Fetching email one message per request can turn an inbox sweep into hours of quiet bash-loop time. The skill points the agent at the batch pattern instead (list messages with metadata in one request, page through), so a bulk scan takes minutes. It triggers on its own once the job looks like bulk mail.

**image-generation** loads when an agent generates or edits images: posters, diagrams, remixed photos. The agent sends its request to the provider with a placeholder credential, and the platform stamps the real key on the way out, so no key lands in the sandbox and usage bills to your credit balance. The skill covers OpenAI's Images API and Google's Gemini image models (Nano Banana), with working examples for both.

None of these skills need installation or configuration. They live in the agent's repository under `skills/` and can be read or customized like any other skill.

## How skills get used

When a task matches a skill's description, the agent loads that skill and follows it. To teach a new skill, describe the workflow in conversation and ask the agent to save it. You can also give it reference material for the skill to draw on.

```text theme={null}
Whenever I ask you to publish the weekly report: pull the numbers from
the metrics sheet, format them in our template, get my sign-off, then
post it to #team. Save that as a skill so I don't have to spell it out
every week.
```

> **[Automate it on a schedule](/docs/agents/routines)**
>
> Pair a skill with a recurring run to automate a whole workflow.

> **[Skills live in the agent repository](/docs/agents/repo)**
>
> Each skill is a folder in the agent's git repository. Clone it to read, edit, or add skills directly.


# Training
Source: https://getskydive.com/docs/agents/training

Review an agent's work, and send it away to study

You train an agent two ways: by **reviewing its work before it acts**, and by **sending it away to study** something too large for one conversation. Both write back to the same [memory](/docs/agents/memory), so what it learns sticks.

## Reviewing before it acts

Training mode lets you check an agent's output before it goes out. Use it while you build trust in a new agent, dial in a workflow, or work on something where a mistake would be costly.

In training mode, the agent presents its work as **artifacts** for review instead of sending or executing directly. You see exactly what it intends to do. You approve or revise it. Only then does it go out.

### When to use it

> **A new agent**
>
> Watch how it handles real tasks before you let it act unsupervised.

> **A new workflow**
>
> Validate a fresh skill or process end to end before trusting it.

> **High-stakes output**
>
> Customer-facing replies, anything irreversible, anything you want a human gate on.

> **Tuning the persona**
>
> See how persona and instruction changes land before they reach anyone.

### The flow

Turn it on by asking, in plain language:

```text theme={null}
Go into training mode. I want to review your replies before they go
out while we dial in the support tone.
```

**1. Turn on training mode**

Ask the agent to enter training mode in your conversation. It applies to that agent until you turn it back off.

**2. The agent drafts**

It does the work and presents the result as an artifact instead of acting.

**3. You review**

Approve it, edit it, or send it back with feedback.

**4. Graduate**

Once the agent consistently gets it right, turn training mode off and let it run on its own.

> [!TIP]
>
> Pair training mode with the [feed](/docs/workspaces/feed): review what the agent produced in the moment, then audit the trail afterward.

## Sending it away to study

Some things take real study: your communication style, the shape of a project, your key relationships, a body of reference material. A normal conversation is too small a window. In a **deep learn** session the agent studies in the background and comes back having absorbed it.

> **Learn your style**
>
> How you write and communicate, so its output sounds like you.

> **Learn a project**
>
> The context, decisions, and players behind a body of work.

> **Learn your world**
>
> Key relationships, accounts, and the things you care about most.

> **Learn a domain**
>
> Reference material and documentation it should master.

### How a session starts

A session begins one of two ways:

* **You ask.** Tell the agent to get up to speed on something: "spend some time learning our brand voice," "get to know my top accounts."
* **The agent proposes one.** When a task needs far more context than a chat holds, the agent suggests a session.

Either way, **you confirm before it runs**. A long session never starts without your go-ahead.

Once confirmed, the session runs in the background. You do not have to stay in the conversation. The agent works through the material on its own and reports back with a summary of what it learned.

### Scoping a session

A good session starts with a clear goal. Before it begins, the agent works out with you:

* **What to focus on**: the specific area to study.
* **What "good" looks like**: concrete outcomes ("you should know my top five accounts," "match my email tone").
* **What to study**: the people, projects, channels, or docs that matter.
* **What it should handle afterward**: the tasks the learning should enable.

With the scope set, the agent runs the session and writes what it learns to its [memory](/docs/agents/memory). The new context applies to every future conversation.

> [!TIP]
>
> Deep learn pays off most for an agent you will rely on long-term, not a one-off task.

> **[Where the learning lives](/docs/agents/memory)**
>
> Everything training and study produce is written to durable memory.

> **[Shape behavior as you review](/docs/agents/persona)**
>
> Fold what you learn in training back into the agent's persona.


# Models for Google-connected agents
Source: https://getskydive.com/docs/agents/trusted-models

How Skydive protects Google data while keeping model choice open

Google-connected agents can use any model that Skydive can route without retaining your data or using it for training.

## Why model choice is limited

Google's [API Services User Data Policy](https://developers.google.com/terms/api-services-user-data-policy) limits how apps may use and transfer data from Google APIs.

Skydive never uses Google data to train a model. It only uses model routes that do not retain Google data or use it for training.

The restriction turns on when an agent connects a Google service through Skydive's OAuth app.

It applies to Gmail, Drive, Calendar, Docs, Sheets, Slides, Contacts, Tasks, and Forms.

## Which models are available

The agent's model picker shows models that Skydive can serve through a zero-data-retention route.

The set can include several providers. It may change as routes and provider terms change.

A model without a qualifying route stays visible but disabled. This prevents a model switch, subagent, or scheduled run from sending Google data through a route that retains it.

## Use a different model

Choose one of these options:

* Pick another available model from the agent's model picker.
* Disconnect every Google OAuth integration from that agent to remove the restriction.
* Use a separate agent without a Google connection.

> [!NOTE]
>
> Disconnecting Google removes that agent's access to the connected Google data. It does not affect other agents.

> **[Choosing a model](/docs/account/models)**
>
> Learn how model selection works and how to match a model to the task.


# Version history
Source: https://getskydive.com/docs/agents/version-history

Browse your agent's past states and restore any of them

Every change an agent makes to itself gets committed to its [git repository](/docs/agents/repo): persona edits, new tools, updated memory, MCP connections. Version history is the timeline of those commits. You can browse it and restore any past state with one click.

## Where to find it

Open the agent's profile page and click the **···** menu, then **Computer**. The **Version History** tab is always available there. You can also reach it directly from the **···** menu via **Manage version history**.

## Reading the timeline

Each row is one commit: what changed, when, and who made it. The newest commit is always marked **Current**. Older rows show a **Restore** button.

The agent makes most commits itself as it works: saving what it learned, wiring up a new tool, updating its persona. When you edit the agent's settings directly, your name appears as the author.

If the agent hasn't made any changes yet, the list shows a single **Initialized** entry with the agent's creation date.

## Restoring a past state

Click **Restore** on any row to roll the agent back to that commit. The agent's repo reverts to that point and its sandbox rebuilds from the result. The timeline keeps polling until the build settles, then updates to show the restored commit as Current.

> [!WARNING]
>
> Restoring replaces the agent's current code, persona file, and tool configuration with the older version. The restore itself is recorded as a new commit, so you can restore again to any point in the list if you change your mind.

Conversations, memory files, and routines are not affected by a restore. Only the repo contents change.

## Cloning the repo

If you have edit access to the agent, a **Copy clone URL** button appears next to the timeline header. It mints a long-lived git token and copies the full clone URL to your clipboard.

```bash theme={null}
git clone <paste the URL here>
```

Editors get a read/write token. If you can view the agent but not edit it, the clone URL is read-only instead.

> [!TIP]
>
> The clone URL is the same repo the agent commits to. You can inspect memory files, review tool definitions, and push changes that the agent will pick up on its next run.

> **[Agent repository](/docs/agents/repo)**
>
> What's inside the repo and how the agent uses it.


# Create an agent
Source: https://getskydive.com/docs/api-reference/agents/create-an-agent

/api-reference/openapi.json post /agents
Creates an agent in your workspace and returns it. The agent is immediately usable: its sandbox, git workspace, and vault are provisioned on first contact. Requires an account-level key with the `edit` scope; agent-bound keys cannot create agents. Subject to your plan's agent quota and a creation rate limit (`429`).



# Get an agent
Source: https://getskydive.com/docs/api-reference/agents/get-an-agent

/api-reference/openapi.json get /agents/{id}
Returns a single agent by id. Responds `404` when the agent does not exist or is not visible to you; the two cases are deliberately indistinguishable.



# List agents
Source: https://getskydive.com/docs/api-reference/agents/list-agents

/api-reference/openapi.json get /agents
Returns the agents in your workspace that you can view, newest first. Use `limit` and the returned `nextCursor` to paginate; a `null` cursor means the last page. Requires an account-level key: agent-bound keys cannot enumerate the workspace.



# Suggest a name for a new agent
Source: https://getskydive.com/docs/api-reference/agents/suggest-a-name-for-a-new-agent

/api-reference/openapi.json get /agents/suggest
Returns a suggested name and avatar for an agent you are about to create, drawn from the workspace's avatar presets. Use it to prefill a creation form. Returns `{ suggestion: null }` when the workspace has no presets.



# Create an API key for an agent
Source: https://getskydive.com/docs/api-reference/api-keys/create-an-api-key-for-an-agent

/api-reference/openapi.json post /agents/{agentId}/api-keys
Creates a `use`-scoped API key bound to this agent. The response is the only time the full `key` value is returned; Skydive stores a hash and cannot recover it. Requires edit access to the agent and a key with the `edit` scope.



# List an agent's API keys
Source: https://getskydive.com/docs/api-reference/api-keys/list-an-agents-api-keys

/api-reference/openapi.json get /agents/{agentId}/api-keys
Returns metadata for an agent's active (non-revoked) API keys: name, display prefix, creation and last-use timestamps. The secret key value is never included; only the first characters are shown, enough to tell keys apart. You must have edit access to the agent.



# Revoke an API key
Source: https://getskydive.com/docs/api-reference/api-keys/revoke-an-api-key

/api-reference/openapi.json delete /agents/{agentId}/api-keys/{id}
Permanently revokes an API key. Requests using it fail with `401` immediately; revocation cannot be undone. The key must belong to the agent in the path. Requires edit access to the agent and a key with the `edit` scope.



# agent
Source: https://getskydive.com/docs/api-reference/cli/agent

Create and manage other agents

`platform agent` creates and manages other agents in your [workspace](/docs/workspaces/overview). A new agent is a peer with its own sandbox, memory, tools, and channels.

| Subcommand | What it does |
| - | - |
| `agent list` | List other agents in the workspace, with roles and descriptions. |
| `agent create [name]` | Create a new agent. Optional `--template`, `--slug`, `--title`, `--description`, `--model`, `--soul`. |
| `agent archive <agent-id>` | Soft-delete an agent you can edit (never itself). |
| `agent unarchive <agent-id>` | Restore an archived agent. |
| `agent update-self` | Update this agent's name, slug, or metadata. A rename propagates automatically to connected channels (Slack display name, etc.). |
| `agent repo [agent-id]` | Read agent repos: every other agent in the organization (or one agent by id), each with its soul.md. Any agent in the org can call this. |
| `agent template` | Browse the curated template catalog (`template list`, and related). |

Archiving removes the agent from the workspace and stops it; `unarchive` brings it back.

## Example

```bash theme={null}
platform agent list
```

```bash theme={null}
platform agent create "Research assistant" \
  --description "Summarizes papers and files notes in Drive"
```

```bash theme={null}
platform agent create "Support Triage" \
  --description "First-line customer support" \
  --soul "You handle support tickets. Triage by severity, draft replies for common questions, and escalate anything urgent."
```

```bash theme={null}
platform agent create --template some-template-slug
```

`--description` is the profile card shown in the workspace directory. `--soul` seeds the persona file (`soul.md`) the agent actually loads as its identity. If you only pass `--description`, the agent boots without a persona. Pass `--soul` when you want a specific voice, role, or set of working principles baked in from the first run.

See [Creating agents](/docs/agents/creating-agents).


# ask
Source: https://getskydive.com/docs/api-reference/cli/ask

Ask the user a multiple-choice question

`platform ask` posts one or more multiple-choice questions to the person in the current conversation as an interactive picker. Options are lettered A, B, C, D, and the picker renders natively on each channel: the web app and CLI show a picker, Slack shows radio buttons, and iMessage shows a plain-text question the user answers by texting back a letter (or their own words).

The call is **non-blocking**: it posts the question and returns immediately. The agent ends its turn, and the user's choice arrives as their next message. There is no polling or waiting.

Use it when the agent genuinely needs the user to choose: a real fork in direction, an ambiguous requirement, or a naming or product call that is theirs to make. It is not for asking permission to do routine, reversible work the agent can just do.

| Flag / arg | What it does |
| - | - |
| `[questions]` | JSON array of 1 to 4 questions. Omit to read `--file` or stdin. |
| `--file <path>` | Read the questions JSON from a file. |

Each question in the array:

| Field | What it does |
| - | - |
| `question` | A short, one-line question ending in `?`. |
| `options` | 2 to 4 choices, each `{ "label": "..." }`. An "Other" option is always offered automatically, so do not add one. |
| `multiSelect` | Optional. `true` lets the user pick more than one (web and iMessage only). |

Slack and iMessage render one question at a time: pass more than one and they fall back to plain text, so on those channels ask a single question. On email and cron there is no picker to render, and the command tells the agent to ask in plain text instead.

## Ask with a default, then proceed

When the agent hits a decision point but could reasonably pick a sensible default, it doesn't have to block the thread waiting for a reply. The pattern: ask with `platform ask`, name the default it will take and roughly when, self-schedule a one-shot wake a few minutes out with `platform automation create --at`, then end its turn.

If you reply before the wake fires, the agent handles your choice and the scheduled run no-ops. If you don't, the wake fires, checks the originating conversation for a reply, and if none has arrived, carries out the stated default.

```bash theme={null}
# Ask, name the default, schedule a fallback
CONV_ID=$(platform conversations list --json | jq -r '.[0].id')
platform ask '[
  {
    "question": "Which format should I use for the export?",
    "options": [
      {"label": "CSV (default in 3 min)"},
      {"label": "JSON"},
      {"label": "Markdown"}
    ]
  }
]'

# Schedule the fallback run
platform automation create \
  --at "$(date -d '+3 minutes' --iso-8601=seconds)" \
  --title "Export format fallback" \
  --summary "Carries out the default CSV export if the user has not replied." \
  "Check conversation $CONV_ID for a reply to the export format question. If none, post back into that conversation and proceed with CSV."
```

Use this only when a default is genuinely reasonable and reversible. For choices that are truly the user's to own, product direction, or anything hard to undo, ask without a scheduled default and wait.

## Example

```bash theme={null}
platform ask '[
  {
    "question": "Which date library should I use?",
    "options": [
      {"label": "date-fns (Recommended)"},
      {"label": "Day.js"},
      {"label": "Luxon"}
    ]
  }
]'
```


# auth
Source: https://getskydive.com/docs/api-reference/cli/auth

Connect external services

`platform auth` sets up and manages authentication for the services an agent uses. Once connected, requests are authenticated automatically through the [proxy](/docs/integrations/permissions); the agent never handles raw tokens.

| Subcommand | What it does |
| - | - |
| `auth connect <service>` | Connect a published integration (OAuth or API key). Renders a Connect card; with no service argument, the best method is picked for you. |
| `auth secret <ServiceName>` | Ask the user for an API key or token and render a Connect card. Intended for services that are not in the catalog; use `auth connect` for catalog services. Simpler than composing a `configure` JSON blob. |
| `auth configure` | Request authentication for an external service (`--config` JSON, advanced). |
| `auth mcp <url>` | Start OAuth for a remote MCP server (renders a Connect card). Grant level is whatever the MCP app requests, not a per-item picker. |
| `auth mark-verified <service>` | Record that a catalog integration's real setup has been completed and verified. |
| `auth reconnect <service>` | Re-run consent for a connected integration without deleting it. Use it to renew or widen an existing grant. |
| `auth disconnect <service>` | Disconnect this agent from a published integration. |
| `auth list` | List all configured auth credentials. |
| `auth enable <credential>` | Enable an auth credential. |
| `auth disable <credential>` | Disable an auth credential. |
| `auth delete <credential>` | Delete a credential and its proxy rules. |
| `auth oauth-providers` | List available OAuth integrations (use `--json` for scopes and suggested rules). |
| `auth catalog` | List published integrations and their connect methods. |
| `auth show <service>` | Show one published integration and its connect methods. |

### Auth types (prefer in this order)

1. **github\_app** - dedicated GitHub App bot identity for the agent
2. **oauth** - one-click provider connect (`integrationKey` + `scopes` + proxy `rules`)
3. **secret** - user-pasted API key, injected by the proxy on matching domains

For CLIs that require an env var before they call the network (for example `gh`), export a sentinel like `GH_TOKEN=sentinel`. The proxy replaces it on the wire.

## Collecting a secret key

When a user offers you an API key or token, use `auth secret`. It renders the same Connect card as `auth configure` but requires just a service name and the API host. No JSON config to compose.

```bash theme={null}
# Custom header and format
platform auth secret Brandfetch --domain api.brandfetch.io --header x-api-key --format '{{token}}'

# Standard Bearer token (default)
platform auth secret OpenAI --domain api.openai.com

# With guidance on where to create the key
platform auth secret Stripe --domain api.stripe.com \
  --instructions 'Create a restricted key at https://dashboard.stripe.com/apikeys'

# Key sent as a URL query param instead of a header
platform auth secret Maps --domain maps.googleapis.com --query-param key
```

The user pastes the key into the card's secure input. It is stored and injected by the proxy on matching requests. The agent never sees or handles the raw value.

After the user submits, test the key against the service by leaving the auth header unset or setting it to `sentinel`. The proxy fills in the real value.

**For catalog services**, use `platform auth connect <service>` instead. The published method supplies the domain and header automatically.

### `auth secret` flags

| Flag | Default | What it does |
| - | - | - |
| `--domain <host>` | required | API hostname to inject the key into. Repeat for more than one host. |
| `--header <name>` | `Authorization` | HTTP header the key is sent in. |
| `--format <template>` | `Bearer {{token}}` (Authorization) or `{{token}}` | Value template. Use `{{token}}` for the key. |
| `--query-param <name>` | none | Send the key as a URL query param instead of a header. Mutually exclusive with `--header`. |
| `--name <slug>` | derived from service name | Credential name (lowercase, underscores). Auto-derived when omitted. |
| `--instructions <text>` | none | Markdown shown on the card: why you need the key and where to create it. |
| `--override` | false | Replace an existing credential with the same name that has not yet been connected. |

## MCP Connect vs OAuth configure

Two connect paths look similar in chat but can grant different scopes:

* **`auth mcp <url>`** opens the MCP server's Connect card. Access is granted at the level the **MCP app** requests (workspace or account). There is no per-page or per-record picker on this path, even when the same provider offers one elsewhere.
* **`auth configure` with `authType: oauth`** is the normal provider OAuth flow. Some providers (Notion is the usual example) show a **per-item picker** on the consent screen for that path only. Empty results after connect often mean the item was not shared on that picker, not that it is missing.

If an agent says an MCP Connect will ask you to pick pages or records, that is the wrong path. That picker only appears on `auth configure` OAuth.

## Example (`auth configure`, advanced)

`auth configure` accepts a full JSON config when you need precise control over rules, scopes, or auth type. For a simple API key, prefer `auth secret` above.

```bash theme={null}
platform auth configure --config "{
  \"service\": \"Example API\",
  \"authType\": \"secret\",
  \"credentialName\": \"example_api_key\",
  \"instructions\": \"Paste an API key from Example > Settings > API.\",
  \"rules\": [{
    \"domain\": \"api.example.com\",
    \"header\": \"Authorization\",
    \"format\": \"Bearer {{token}}\"
  }]
}"
```

## Example (list providers)

```bash theme={null}
platform auth oauth-providers --json
platform auth list
```

See [Connecting services](/docs/integrations/oauth) and [Supported integrations](/docs/integrations/catalog).


# automation
Source: https://getskydive.com/docs/api-reference/cli/automation

Run on a schedule

`platform automation` creates and manages an agent's routines: recurring or one-shot work that fires with no human in the loop. Alias: `scheduled-run`.

| Subcommand | What it does |
| - | - |
| `automation list` | List this agent's schedules with their last run outcome. |
| `automation show <id>` | Show one schedule (next fire, last run). |
| `automation history <id>` | List recent firings for a schedule: when it fired, status, duration, conversation id, and any error. Alias: `runs`. |
| `automation run <id>` | Fire a schedule once, right now, without waiting for its cron. |
| `automation create <description>` | Create a schedule. Requires `--cron` or `--at`, plus `--title` and `--summary`. |
| `automation update <id>` | Update description, title, summary, cron, time, timezone, or model. |
| `automation pause <id>` | Pause (resumable). |
| `automation resume <id>` | Resume a paused schedule. |
| `automation cancel <id>` | Cancel permanently. |

### Required on create

| Flag | Role |
| - | - |
| `<description>` | The instruction the agent runs when the schedule fires (not just a label). |
| `--title` | Short name in the UI (a few words). |
| `--summary` | One or two plain sentences of what it does (UI). No em dashes. |
| `--cron` or `--at` | Recurring cron expression, or a single ISO-8601 time. |

Optional: `--tz` for human-meaningful cron (defaults to the calling user timezone, or UTC). On `update`, `--tz` works on its own to re-stamp a recurring schedule's timezone while keeping its stored cron expression. It has nothing to change on a one-off: the `--at` instant is already fixed, so moving a one-off means passing `--at` again (with `--tz` if needed).

On `update`, pass `--model <name>` to change which model a schedule fires on, or `--clear-model` to fall back to the agent's default. Pass neither to leave the model unchanged.

Firings run with no one watching, so put every hard fact the run needs (targets, accounts, IDs, what done looks like) in the description.

## Example

```bash theme={null}
platform automation create \
  --cron "0 9 * * 1-5" \
  --tz America/Los_Angeles \
  --title "Weekday standup digest" \
  --summary "Compiles open PRs and posts a short standup note every weekday at 9am Pacific." \
  "Pull open PRs for Create-Inc/skydive-docs since yesterday. Post a short standup bullet list in Slack #docs."
```

```bash theme={null}
platform automation create \
  --at 2026-08-01T17:00:00-07:00 \
  --title "Launch day reminder" \
  --summary "One-shot reminder on launch day at 5pm Pacific." \
  "Remind the team in #launches that GA is today. Include the changelog link."
```

Fire a schedule immediately to test it before its first unattended run. It goes through the normal run path (executor lock, overlap policy), so it is skipped if a run is already in flight.

```bash theme={null}
platform automation run auto_01h9
```

To check what a schedule actually did, use `automation history`. It shows each firing's status, how long it took, and the conversation id the run opened (so you can pull the transcript with `platform conversations show`):

```bash theme={null}
# Last 20 firings (default)
platform automation history auto_01h9

# Last 50
platform automation history auto_01h9 --limit 50

# JSON output
platform automation history auto_01h9 --json
```

A `skipped` firing means a run was already in flight when the cron fired. A `failed` firing includes the error reason, and is retried automatically: a failed firing fires again within a minute, up to 3 attempts per fire window. `automation history` marks a retry with a `retry: attempt N` line under the firing, and `--json` includes an `attempt` field, so you can tell a failure that recovered from a terminal one.

`automation list` shows a LAST RUN column for each schedule: the outcome of its most recent firing (e.g. `skipped (billing_blocked)`, `failed`, `completed`). One-shot schedules that were skipped due to billing show `completed` in the job lifecycle but `skipped (billing_blocked)` as the last run, so you can tell at a glance whether the work actually ran.

See [Routines](/docs/agents/routines).


# billing
Source: https://getskydive.com/docs/api-reference/cli/billing

Check your workspace's billing state from inside an agent

`platform billing status` lets an agent report on the workspace's credit balance, surface billing-related run blocks to users, and hand off the right link to resolve them. It requires the caller to be an authenticated workspace member.

| Command | What it does |
| - | - |
| `billing status` | Print the workspace's billing state: balance, plan, block reason (if any), and checkout links. |
| `billing status --json` | Same, as structured JSON. |

## When to use it

Most agents don't need this. Reach for it when:

* A run was blocked mid-flight and you want to tell the user why and where to go.
* You're building a billing-aware flow (a scheduled job that checks balance, a "before I start, are we good?" preflight).
* A user asks about credit balance or billing status and you want to give them the live number rather than the docs page.

The links returned by `billing status` are workspace-scoped and member-only. Don't send them to external or anonymous users. They link directly into `/settings/billing` on the right sub-page.

## Example

```bash theme={null}
platform billing status

# --- output ---
# Plan: Starter
# Balance: $14.32 remaining
# Status: active
# Links: buy credit, auto top-up, manage plan
```

```bash theme={null}
platform billing status --json
```

Returns fields including `balance`, `plan`, `blockReason`, and a `links` object with per-action deep links.

## Track spending with `platform usage`

`platform usage` shows what this agent has spent, estimated per UTC day and priced from the workspace's effective rates. Estimates are not the actual credit draw-down; `billing status` shows the real balance.

| Command | What it does |
| - | - |
| `platform usage` | Daily spend for the last 7 days, grouped by usage category (AI, compute, and so on). |
| `platform usage --days <1-31>` | Include more or fewer days. |
| `platform usage --group-by model` | Group the daily cost by model instead of category. |
| `platform usage --json` | Print the raw payload as JSON. |

```bash theme={null}
platform usage --days 2

# --- output ---
# Usage for this agent: 2 days (2026-09-28 to 2026-09-29, UTC days), grouped by category.
# Estimated total: $0.96 (AI $0.76, Compute $0.20)
#
# By day (UTC, series with zero cost omitted):
#   2026-09-28  AI $0.55, Compute $0.20
#   2026-09-29  AI $0.21
```

## Block reasons

When `blockReason` is present, the workspace is billing-blocked and runs may not start. Common values and what to tell the user:

| blockReason | What it means | Where to send the user |
| - | - | - |
| `insufficient_balance` | Credit ran out | Buy credit link |
| `payment_failed` | The last payment didn't go through, so the card on file needs updating | Update card link |
| `hard_spend_limit_reached` | Org hit an admin-set spend cap | Contact the workspace admin |

> [!NOTE]
>
> `billing status` is gated to authenticated workspace members. External senders and agents outside the workspace get a 403. Billing links are member-only: don't forward them to end users who aren't members of your workspace.

See [Pricing](/docs/account/pricing) and [Usage and billing](/docs/account/usage) for the user-facing billing pages.


# builds
Source: https://getskydive.com/docs/api-reference/cli/builds

Sandbox template build status

Every time the agent code changes, Skydive rebuilds its sandbox template in the background. `platform builds` shows whether the current head is built (so a restart boots the latest code) and surfaces full build errors when something fails.

Builds are triggered by pushing to the [agent's git repository](/docs/agents/repo). The `Dockerfile` in that repo defines what gets baked into the image.

| Subcommand | What it does |
| - | - |
| `builds current` | What this sandbox is actually running right now (build pin, template, built-at), compared with the workspace head. Flags stale or fallback images. |
| `builds list` | Recent build attempts: whether current head is built, plus failures. |
| `builds show [buildId]` | Full detail for one attempt (complete error text). Omitting the id defaults to the most recent failure. |
| `builds wait` | Block until the current head builds successfully (or fail / time out). Useful right after a push. |

## Example

```bash theme={null}
platform builds list
platform builds current
platform builds show
platform builds wait
```

After `git push`, `builds wait` is the quick check that the new template is ready before you depend on it.


# channel
Source: https://getskydive.com/docs/api-reference/cli/channel

Channel context

`platform channel` reads context about the conversation the agent is in right now and controls how the agent responds in it.

| Subcommand | What it does |
| - | - |
| `channel info` | Current channel context: channel type, ids, thread, and participants where available. |
| `channel sender` | Resolve who the agent is talking to: the verified email and linked channel identities of the person who started the run, plus the other people in the conversation. Used instead of asking someone to identify themselves. |
| `channel suppress-reply` | Suppress the automatic reply for this message, for when the agent handles the response itself or a group message isn't for it. |
| `channel propose <channel>` | Propose that the user connect a channel (Slack, email, iMessage), returning a card they can act on. |

## Example

```bash theme={null}
platform channel info --json
```


# compute
Source: https://getskydive.com/docs/api-reference/cli/compute

View or request a change to the sandbox compute tier

`platform compute` lets an agent check its current sandbox tier and ask for a bigger one when a job is genuinely compute-bound. An agent cannot resize itself: a request posts an Approve/Deny card, and a person who can edit the agent decides. On approval the tier is raised and the sandbox restarts at the new size.

| Subcommand | What it does |
| - | - |
| `compute status` | Current tier and which higher tiers can be requested. |
| `compute request` | Ask for a bigger sandbox (`--tier` required, `--reason` shown to the approver). |
| `compute cancel` | Withdraw a pending request. Only one can be open at a time. |
| `compute metrics` | Show this sandbox's CPU, memory, and volume usage over time as sparklines (`--window-minutes`, `--json`). Available where the workspace has it turned on. |

Tiers (ascending): `xsmall` \< `small` \< `medium` \< `large` \< `xlarge` \< `xxlarge` \< `xxxlarge`. You can only request a tier above the current one. Requests need an active conversation turn (routines cannot request compute).

## Example

```bash theme={null}
platform compute status
```

```bash theme={null}
platform compute request --tier large \
  --reason "The test suite OOMs at 2GB; it needs about 6GB to run green."
```

```bash theme={null}
platform compute cancel
```

```bash theme={null}
platform compute metrics --window-minutes 120
```


# computer
Source: https://getskydive.com/docs/api-reference/cli/computer

Browser and file tools

`platform computer` covers the agent's Computer: its [browser](/docs/capabilities/browser) and the [Files](/docs/capabilities/files) area (`~/Documents`). Both subgroups produce share links so the agent can hand a browser session or a file to you on channels without a side panel.

| Subcommand | What it does |
| - | - |
| `computer browser share-link` | Link to watch and control the agent's live browser (sign-in, CAPTCHA, 2FA). |
| `computer files share-link <path>` | Link to view and download a file from `~/Documents`. |
| `computer ports list` | Ports this sandbox is listening on, with each one's public URL. |

In web chat you already see the browser and files in the Computer panel, so these links are mainly for Slack, email, and text.

## Example

```bash theme={null}
platform computer browser share-link
platform computer files share-link ~/Documents/report.pdf
platform computer ports list
```


# conversations
Source: https://getskydive.com/docs/api-reference/cli/conversations

Read past conversations and start new ones, and post into existing ones

`platform conversations` reads this agent's past conversations and messages on the active branch, starts new ones, and posts into existing ones to steer a running subagent or queue work for its next turn.

| Subcommand | What it does |
| - | - |
| `conversations create` | Open a new web conversation with an opening message the agent acts on immediately. |
| `conversations post <id>` | Post a message into an existing conversation and dispatch a run on it. |
| `conversations list` | List recent conversations. Pass `--owner-only` to scope the results to conversations where the agent owner appears as a participant, or `--mode a2a` to list only agent-to-agent conversations. |
| `conversations show <id>` | Show one conversation. |
| `conversations search <query>` | Search across past conversations. |

## Starting a new conversation

Use `conversations create` when the agent needs to reach out and start a fresh thread rather than reply in an existing one. The opening message is dispatched as a run immediately.

```bash theme={null}
platform conversations create --message "Your weekly report is ready."
platform conversations create --message "Found 3 failing tests." --title "Test failures: 2026-08-04"
```

Returns the new `conversationId` and `runId`. Pass `--json` for machine-readable output.

## Posting into an existing conversation

Use `conversations post` to send a message into a conversation the agent already participates in. A run is dispatched immediately.

```bash theme={null}
# Positional form (recommended)
platform conversations post <id> "Prioritize the billing module first."

# Flag form (also accepted)
platform conversations post <id> --message "Prioritize the billing module first."
```

The message flag accepts the same [`@file` shortcut as `platform post`](/docs/api-reference/cli/post): `--message @notes.txt` sends the file's contents.

The behavior depends on whether the target is mid-run or idle:

* **Running:** the message steers the run in progress, like a mid-turn steer directive.
* **Idle:** the message queues for the next run.

The most common use is steering a subagent. The `subagent` tool returns each child's `conversationId`; pass that id to `conversations post` to send it instructions without opening a new thread.

```bash theme={null}
# Steer a child agent mid-run
platform conversations post 019abc... "Skip the linting step, just run tests."
```

You can only post into conversations you participate in. Posting into another agent's conversation returns 404.

## Reading conversations

```bash theme={null}
platform conversations list
platform conversations search "billing path"
platform conversations show <id>
```

### Filtering by owner

Pass `--owner-only` to limit results to conversations where the agent's owner appears as a participant. This is useful in proactive routines that should target conversations the owner is actually part of, rather than every conversation the agent has ever joined.

```bash theme={null}
platform conversations list --owner-only
```

The owner match covers web chat, linked Slack identities, and linked iMessage identities. The result also includes channel, audience, and recent-activity metadata to help your agent decide which conversation is the right one to act in.

### Filtering agent-to-agent and by participant

Two more filters narrow a listing when a routine needs a specific kind of thread:

```bash theme={null}
# Only agent-to-agent conversations (no human participant)
platform conversations list --mode a2a

# Only conversations where a specific human is an active participant
platform conversations list --human-participant <user-id>
```

`--human-participant` also accepts `sender` (the human who started the current turn) or `owner` (this agent's owner) instead of a user id.


# db
Source: https://getskydive.com/docs/api-reference/cli/db

Provision and manage Postgres databases

`platform db` provisions managed Postgres databases (hosted on Neon) for the apps an agent builds, and attaches them so the connection string is available as an environment variable. Databases are shared at the workspace level: any agent in the workspace can attach one. See [Databases](/docs/capabilities/storage).

| Subcommand | What it does | | |
| - | - | - | - |
| `db list` | List workspace databases and how this agent has each attached. | | |
| `db create <slug>` | Provision a new database. Optional `--attach <ENV_VAR>`, \`--size small | medium | large\`. |
| `db attach <slug> <ENV_VAR>` | Attach a workspace database to this agent as a secret env var. | | |
| `db detach <slug>` | Detach from this agent (database itself stays). | | |
| `db archive <slug>` | Soft-delete workspace-wide (data kept, slug freed). | | |

Slug is lowercase (for example `default`). Size defaults to `small`.

## Example

```bash theme={null}
platform db create default --attach DATABASE_URL --size small
```

```bash theme={null}
platform db list
platform db attach analytics DATABASE_URL
```

Launch processes that need the URL through `platform secrets run` so the env var is injected:

```bash theme={null}
platform secrets run -- node server.js
```


# deep-learn
Source: https://getskydive.com/docs/api-reference/cli/deep-learn

Background learning sessions

`platform deep-learn` starts a background learning session when the agent needs far more context than one conversation holds: a person's writing voice, a codebase, a book of accounts. The agent scopes the goal with you first, then proposes the session.

| Subcommand | What it does |
| - | - |
| `deep-learn propose` | Propose a session for the builder to confirm. Takes the scope as guidance. |
| `deep-learn list` | List this agent's sessions. |
| `deep-learn status` | Check one session by id. |
| `deep-learn done` | Signal that the session is complete. |
| `deep-learn test` | Test the agent against a task using only what it has persisted, to see whether the learning stuck. |

A session runs in the background, so the agent stays available while it learns. `deep-learn test` is the honest check: it withholds the source material and asks the agent to work from memory alone.

## Example

```bash theme={null}
platform deep-learn propose --guidance "Learn how I write: read my last 200 posts and match my voice."
platform deep-learn list
```

See [Deep learn](/docs/agents/training#sending-it-away-to-study).


# email
Source: https://getskydive.com/docs/api-reference/cli/email

Email channel tools

`platform email` works with the agent's email channel: reading threads, sending, replying, and pulling attachments out of a message. For sending, you can use either `platform email send` or `platform post email`; both reach the same delivery path.

An agent needs a connected inbox before any of this works. Until then `platform email` has nothing to act on, and the agent has no address to hand out. See [Email](/docs/channels/email) for connecting one.

| Subcommand | What it does |
| - | - |
| `email self` | Show the agent's own email address and display name. |
| `email threads` | List recent threads (`--limit`, max 100; `--page-token` to page). |
| `email thread <threadId>` | Show every message in one thread. |
| `email send` | Compose and send a new email. |
| `email sent` | List messages this agent has sent (newest first). Useful to confirm a send actually went out. |
| `email reply <messageId> [text]` | Reply to the sender. `--cc` loops others back in. |
| `email reply-all <messageId> [text]` | Reply to everyone on the message. |
| `email forward <messageId>` | Forward to new recipients (`--to`). |
| `email file-download <attachmentId>` | Download an attachment into the working directory (`--output` to rename). |

### Composing

`send`, `reply`, `reply-all` and `forward` all take the body as a positional argument, or read it from stdin when you leave it off.

| Flag | Applies to | What it does |
| - | - | - |
| `--to` | `send`, `forward` | Recipients. Required. |
| `--subject` | `send` | Subject line. Required. |
| `--cc` / `--bcc` | `send`, `reply` | Additional recipients. |
| `--html` | `send`, `reply`, `reply-all` | HTML body, sent alongside the plain text. |
| `--attach` | `send`, `reply`, `reply-all` | File paths to attach. |
| `--inline` | `send`, `reply`, `reply-all` | Images to embed. Reference them in `--html` as `<img src="cid:filename">`. |

> [!WARNING]
>
> Write the body as ordinary prose. Apostrophes and quotes are captured verbatim inside the double quotes, so POSIX-escaping them (`I don'\''t`) puts the escape characters into the email itself.

## Examples

```bash theme={null}
# What is my address?
platform email self

# Read the inbox
platform email threads --limit 10
platform email thread <thread-id>
```

```bash theme={null}
# Send, with an attachment
platform email send "Report attached. Numbers are through Friday." \
  --to person@example.com \
  --subject "Weekly numbers" \
  --attach ~/Documents/weekly.pdf
```

```bash theme={null}
# Reply, looping in a colleague
platform email reply <message-id> "Confirmed, shipping Thursday." --cc lead@example.com
```

See [Email](/docs/channels/email).


# feed
Source: https://getskydive.com/docs/api-reference/cli/feed

Post rich cards to the agent feed

`platform feed` publishes cards to the [feed](/docs/workspaces/feed), the shared home timeline where an agent's work shows up for the team.

Most of the time the agent does not need this: conversation runs are summarized into the feed automatically. Post a card when no one is watching (routine, agent-to-agent) or when a result deserves the whole team.

| Subcommand | What it does |
| - | - |
| `feed post [content]` | Publish a card. Pass JSON as a positional arg, `--content-file`, or stdin. For a simple text card, use `--title` and `--body` instead. |
| `feed list` | List this agent's recent cards, newest first. |
| `feed show <activity-id>` | Read back the full stored content of a card by its id. Useful for verifying what was posted. Pass `--json` for the raw render spec. |
| `feed delete <activity-id>` | Retract a card you posted (a card cannot be edited in place, so delete a wrong one and post a corrected one). |

A card is one result, not a form. Lead with the outcome. Prefer stats, charts, and lists over long prose. Use only real numbers and links from the work.

Common block types: `text`, `heading`, `stats`, `chart`, `list`, `keyValue`, `image`, `video`, `button`, `divider`.

The `video` block renders an inline video player. Fields: `url` (required, HTTPS), `caption` (optional, up to 200 characters), and `posterUrl` (optional thumbnail image). Videos that fail to load are hidden automatically.

## Posting a card

Two forms work for `feed post`.

**Simple text card**: `--title` and `--body` together post a heading and a text paragraph. Both flags are required when you use this form.

```bash theme={null}
platform feed post --title "Deploy complete" --body "Merged 3 PRs, all checks passed."
```

**Rich card**: pass a JSON object as the positional arg (or via `--content-file`/stdin) to use the full block vocabulary.

```bash theme={null}
platform feed post '{
  "title": "Signups up 18% week over week",
  "blocks": [
    {"type": "stats", "items": [
      {"label": "Signups", "value": "1,204", "delta": "+18%", "trend": "up"}
    ]},
    {"type": "text", "text": "Organic search led the gain. Paid held flat."},
    {"type": "button", "label": "Open dashboard", "url": "https://example.com/dash"}
  ]
}'
```

You cannot mix both forms in the same call.

```bash theme={null}
platform feed list
```

See [Feed](/docs/workspaces/feed).


# feedback
Source: https://getskydive.com/docs/api-reference/cli/feedback

Report platform problems

`platform feedback` files a structured report when the platform itself gets in the agent's way (bug, missing capability, infra).

| Subcommand | What it does |
| - | - |
| `feedback report [body]` | Send a structured report to the Skydive team. |

## Example

```bash theme={null}
platform feedback report "platform automation update returned 404 for a known id"
```


# imessage
Source: https://getskydive.com/docs/api-reference/cli/imessage

SMS and iMessage tools

`platform imessage` acts in the SMS and iMessage conversation the agent is currently handling. Most subcommands target that conversation, so they don't take a recipient. One exception: `imessage self` works with no active thread. To send a text message you can use either `platform imessage send` or `platform post imessage`; both do the same thing.

| Subcommand | What it does |
| - | - |
| `imessage self` | Print the agent's own iMessage number and how it is provisioned (`dedicated` or `pool`). Works with no active conversation. |
| `imessage send <text>` | Send a text message. |
| `imessage history` | Recent messages in this conversation (`--limit`, `--cursor`). |
| `imessage react <messageId> <reaction>` | React with `love`, `like`, `dislike`, `laugh`, `emphasize`, `question`, or `custom` plus `--emoji`. |
| `imessage unreact <messageId> <reaction>` | Remove a reaction. |
| `imessage send-media <url> [urls..]` | Send images, videos, or files by public HTTPS URL. Several URLs go as one stacked group. |
| `imessage upload [filename] [contentType] [base64Content]` | Send a local file. `--files-json` takes an array of `{filename, contentType, base64Content}` to send several as one group. |
| `imessage send-link <url>` | Send a URL with a rich link preview. |
| `imessage send-voice-memo <url>` | Send audio as a voice memo, with inline playback. |
| `imessage share-contact` | Send the agent's contact card, so the person can save it. |

Replying is usually simpler than `imessage send`: text output from the agent's turn is delivered as the reply automatically. Reach for `send` when a message has to go out mid-turn, and for the media subcommands when the reply is a file rather than words.

## Examples

```bash theme={null}
# React instead of replying
platform imessage react <message-id> like

# Send a chart the agent just made
platform imessage upload chart.png image/png "$(base64 -w0 chart.png)"
```

```bash theme={null}
# Catch up on the thread
platform imessage history --limit 50
```

## Your own number

`platform imessage self` prints the phone number the agent sends from, along with how it is provisioned.

```bash theme={null}
platform imessage self
```

This works even with no active iMessage thread, so an agent can look up its own number at any point in a run. Two provisioning types:

| Type | What it means |
| - | - |
| `dedicated` | The number belongs exclusively to this agent. |
| `pool` | The number is shared across workspaces. The agent still sends and receives on it; it just isn't exclusive. |

An agent on a shared line can explain this to a user: "My number is +1 (xxx) xxx-xxxx. It's a shared line, so you may see it associated with other agents."

Use `platform channel info` while handling an iMessage thread for the thread-scoped view, which also surfaces the number alongside chat id, sender, and participants.

See [SMS and iMessage](/docs/channels/text).


# models
Source: https://getskydive.com/docs/api-reference/cli/models

Browse the model catalog and set which model your agent runs

`platform models` gives an agent access to the full model catalog and lets it manage its own model selection. Use it to browse available models, pick one for delegated work, or switch the model the agent itself runs on.

The catalog shows every model id, provider, cost (USD per one million input and output tokens), context window, reasoning support, and image-output capability, sorted cheapest first. The row marked with `*` is the model an omitted `--model` inherits when delegating.

| Subcommand | What it does |
| - | - |
| `models list` | The full catalog: model id, provider, input/output cost per 1M tokens, context window, reasoning support, and image-output capability, cheapest first. Running `platform models` with no subcommand does the same thing. |
| `models set <model>` | Set this agent's default model. Takes the exact model id from `models list`. Pass `--thinking-level <level>` to pin a reasoning effort at the same time. Pass `--agent <id\|slug\|name>` to set another agent's default model in this workspace instead (staff opt-in: the server rejects the workspace if it is not enabled). |
| `models reset` | Clear the agent's model back to the platform default. Also accepts `--agent <id\|slug\|name>` to clear another agent's default. |

Add `--json` to `models list` to get the catalog as structured JSON for scripting. The `IMAGE` column in the output flags models that can generate images; use it when picking a model for image tasks rather than guessing from the model name.

`models set` also accepts `--conversation` to pin the model for the current conversation only, without changing the agent's default. `models reset --conversation` clears a per-conversation pin.

`--agent` sets a different agent's stored default instead of this agent's. It takes an id, slug, or name from `platform agent list`, and it cannot be combined with `--conversation` or `--thinking-level`: a per-conversation pin on the caller's thread is a different surface. The server only accepts it for workspaces enrolled in staff-level delegation.

## Example

```bash theme={null}
platform models list
platform models list --json
```

Pick a cheap model from the list for routine delegated work, then pass its id when you create or delegate to a subagent:

```bash theme={null}
platform agent create "Research helper" --model <id>
```

An agent can also set its own model at runtime:

```bash theme={null}
# Switch this agent to a faster model for the rest of its runs
platform models set deepseek/deepseek-v4.1-flash

# Pin a different model for just this conversation
platform models set anthropic/claude-fable-5-1 --conversation

# Clear the per-conversation pin
platform models reset --conversation

# Go back to the platform default
platform models reset
```


# notify
Source: https://getskydive.com/docs/api-reference/cli/notify

Raise an in-app notification, or omit --kind to land in the daily digest email

`platform notify` reaches the person in the current conversation. **View** in the web app takes them back here.

Omit `--kind` for a proactive update: something you finished on your own that you'd have sent if you could message them. That lights the bell and groups into their next daily digest email, which is how you reach someone on a channel with no connected messaging surface.

Pass `--kind` only when a human should look **now**:

* **permission**: you need access or approval to continue
* **alert**: something significant to report

Those two light the bell only. They are not emailed.

This is not an activity log. Routine progress belongs in the [feed](/docs/workspaces/feed). Notifications only reach the user on the current conversation. On some Slack, email, or SMS turns with no logged-in web user, the call succeeds quietly and sends nothing.

| Flag / arg | What it does |
| - | - |
| `[body]` | Notification text. Omit to read `--body-file` or stdin. |
| `--kind alert\|permission` | Attention now (bell only). Omit for a proactive update that rides the daily digest. |
| `--body-file <path>` | Read the body from a file. |

## Example

```bash theme={null}
platform notify "Finished the weekly usage recap. Numbers are in the feed."
```

```bash theme={null}
platform notify --kind permission \
  "I need read access to the shared Drive folder to finish the audit."
```

```bash theme={null}
platform notify --kind alert "Weekly usage crossed 80% of the plan include."
```


# portal
Source: https://getskydive.com/docs/api-reference/cli/portal

Reach a machine running Skydive Desktop

`platform portal` reaches a computer running Skydive Desktop after the user grants this agent access. Default deny; consent is per agent and revocable.

> [!NOTE]
>
> `platform portal request` only works when the user is in a web chat conversation or the `skydive chat` CLI. Those are the only surfaces where a Skydive Desktop app or CLI is running locally to handle the approval card. Calling it from Slack, email, iMessage, or a scheduled run returns an error.

| Subcommand | What it does |
| - | - |
| `portal request` | Ask the user to grant Desktop access (approval card). Only works from a web chat or the `skydive chat` CLI, where the user has a Desktop app or CLI running to approve the card. |
| `portal list` | List machines this agent can use. |
| `portal exec` | Run a command on a connected machine. |
| `portal bind` | Tunnel a machine's local port into this sandbox (reach a server on the desktop). |
| `portal expose` | Publish a sandbox port onto the desktop's loopback (the user opens `localhost:<port>` in their own browser). |
| `portal tunnels` | List active port tunnels (`bind` and `expose`) with their live or dead state. |
| `portal unbind <port>` | Close a tunnel opened with `bind` or `expose` on that sandbox-side port. |
| `portal file read` | Print a file from the desktop to stdout. |
| `portal file write` | Write a file to the desktop from a local path, inline content, or stdin. |
| `portal file edit` | Make exact-text replacements in a file on the desktop. |
| `portal import-secrets <path>` | Read a `.env` file from the desktop and import its key-value pairs as agent secrets. |

## File operations

`portal file` reads, writes, and edits files on a connected desktop without hand-rolling `cat`/`base64` pipelines through `exec`. Bytes move as base64 internally, so binary files survive; you never see the encoding.

```bash theme={null}
platform portal file read <path> [--offset N] [--limit N]
platform portal file write <path> [--from <localPath> | --content <str> | --stdin] [--mkdir]
platform portal file edit <path> --old <str> --new <str> [--old <str> --new <str> ...]
platform portal file edit <path> --edits '<json array of {oldText,newText}>'
```

`edit` uses exact-match-first then fuzzy matching (smart quotes, unicode dashes, trailing whitespace). Each `--old` must match exactly once in the file. Multiple edits are matched against the original file and applied together.

`--machine` is optional on all subcommands when exactly one machine is connected.

## Tunneling ports

Two directions, depending on which side hosts the server:

* `portal bind <port>` pulls a port on the desktop into this sandbox, so the agent can reach a server running on the user's machine (a local API, a dev server, a local MCP).
* `portal expose <port>` pushes a port from this sandbox onto the desktop's loopback, so the user opens `localhost:<port>` in their own browser and hits a server the agent is running. Pass `--on <port>` to listen on a different port on the desktop.

`portal tunnels` lists every active tunnel from both commands with an honest live or dead state, so a stale tunnel is visible rather than silently broken. Each row includes a hint for closing the tunnel: `(close with: platform portal unbind <port>)`.

`portal unbind <port>` closes a tunnel on that sandbox-side port immediately. Unbinding a port with no tunnel is a no-op.

```bash theme={null}
platform portal bind 5432                 # reach the desktop's Postgres from the sandbox
platform portal expose 3001               # let the user open the sandbox's web app locally
platform portal expose 3001 --on 8080     # publish it on a different desktop port
platform portal tunnels                   # list active tunnels and their state
platform portal unbind 5432              # close the tunnel on port 5432
```

## Importing secrets from the desktop

`portal import-secrets` reads a `.env` file off a connected desktop and loads its key-value pairs as agent secrets in one step. Secret values are never printed; only names are shown.

```bash theme={null}
platform portal import-secrets ~/.env                     # read and import
platform portal import-secrets ~/.env --dry-run           # preview names without importing
platform portal import-secrets ~/.env --machine mybook    # target a named machine
```

The file is parsed the same way `platform secrets set --env-file` does it: quoted values, `export`-prefixed lines, comments, and `=` in values all work. A missing file, an empty file (no key-value pairs), or a malformed line fails loudly and imports nothing.

## Example

```bash theme={null}
platform portal request
platform portal list
platform portal file read ~/project/config.json
platform portal file edit ~/project/config.json --old '"port": 3000' --new '"port": 8080'
platform portal import-secrets ~/project/.env
```

See [Reaching your computer](/docs/integrations/portal).


# post
Source: https://getskydive.com/docs/api-reference/cli/post

Send a message to Slack, email, or iMessage from the agent's sandbox

`platform post` is a unified posting command. It wraps the same delivery functions as the service-specific commands, so `platform post slack` and `platform slack post` do the same thing. Use whichever reads more naturally in context.

| Subcommand | What it does |
| - | - |
| `post slack` | Post a message to Slack. Defaults to the current conversation's channel and thread. |
| `post email` | Send or reply to an email. |
| `post imessage` | Send a text message in the current iMessage or SMS conversation. |

Every message argument also accepts a file: `--message @report.txt` sends the contents of report.txt rather than the literal text. A leading `@` followed by a path is the trigger. If the path doesn't resolve, the command fails with an error instead of posting the string as-is.

## Slack

```bash theme={null}
# Post in the current channel
platform post slack --message "Deployment complete."

# Post to a specific channel at root (no thread)
platform post slack --channel C123ABC --message "Weekly summary attached."

# Reply in a thread
platform post slack --channel C123ABC --thread 1234567890.123456 --message "Done."
```

All flags from `platform slack post` are available: `--no-thread`, `--force`, `--broadcast`, `--blocks`, `--json`.

## Email

```bash theme={null}
# Send a new email
platform post email --to person@example.com --subject "Update" --message "Here is the report."

# Reply to a message
platform post email --reply-to <messageId> --message "Thanks, confirmed."

# Reply-all
platform post email --reply-to <messageId> --reply-all --message "Looping everyone in."
```

Supports HTML (`--html`), attachments (`--attach`), inline images (`--inline`), CC/BCC, and `--json`. An inbox must be connected before this works. See [Email](/docs/channels/email).

## iMessage

```bash theme={null}
platform post imessage --message "On my way."
```

Targets the iMessage or SMS conversation the agent is currently handling. Pipe text from stdin if you prefer: `echo "Done" | platform post imessage`. See [SMS and iMessage](/docs/channels/text).


# search
Source: https://getskydive.com/docs/api-reference/cli/search

Search and read the web

`platform search` finds and fetches up-to-date information from the web. Prefer this over driving a browser when you only need page text.

| Subcommand | What it does |
| - | - |
| `search query <query>` | Search the web. Optional `--num-results` and `--json`. |
| `search get-contents <url> [url2 ...]` | Fetch the text content of one or more URLs. |

## Example

```bash theme={null}
platform search query "Skydive agent routines" --num-results 5
platform search get-contents https://www.skydive.com/docs/agents/routines
```

See [Web search](/docs/capabilities/web-search).


# secrets
Source: https://getskydive.com/docs/api-reference/cli/secrets

Store API keys securely

`platform secrets` stores values the agent can use without putting them in prompts or git. Prefer [auth](/docs/api-reference/cli/auth) when the service can use OAuth or a proxy rule.

| Subcommand | What it does |
| - | - |
| `secrets list` | List secret names available to this agent. Never prints secret values, including with `--json`. |
| `secrets get <name>` | Read a single secret value (use carefully; prefer `secrets run`). |
| `secrets set` | Store or update a secret. See flags below. |
| `secrets delete <name>` | Remove a secret. |
| `secrets run -- <command>` | Run a command with secrets injected as env vars. |
| `secrets download` | Print all secrets to stdout (use `--plain` for raw values). |

`list` is names only. Use `get` when you truly need a value, or `run` so the value is injected as an environment variable and never lands in the transcript.

## `secrets set`

Passing a secret value as a command-line argument can expose it to other processes reading `/proc/*/cmdline` or `ps`, and it lands in shell history. Two safer input paths are available:

| Flag | What it does |
| - | - |
| `--env-file <path>` | Read `KEY=VALUE` pairs from a `.env` file. Pass `-` to read from stdin. |
| `--stdin` | Read `KEY=VALUE` pairs from stdin, one per line. |

Positional `KEY=VALUE` pairs still work. When multiple sources are given, they merge in order of precedence: env file first, stdin next, then positional arguments. Secret names are printed when set; values are never printed.

```bash theme={null}
# Load an entire .env file
platform secrets set --env-file .env

# Pipe values from stdin
cat .env | platform secrets set --stdin

# Mix: env file + one override on the command line
platform secrets set --env-file .env DATABASE_URL="postgres://..."
```

## Example

```bash theme={null}
platform secrets list
platform secrets run -- node server.js
```


# slack
Source: https://getskydive.com/docs/api-reference/cli/slack

Slack channel tools

`platform slack` reads and posts in connected Slack workspaces. To post a message, you can use either `platform slack post` or the newer `platform post slack`; both interfaces do the same thing.

| Subcommand | What it does |
| - | - |
| `slack self` | Show the bot Slack identity. |
| `slack channels` | List channels the bot can see. Most are not joined yet. Pass `--dms` for DMs and group DMs. |
| `slack join <channel>` | Join a channel. |
| `slack leave <channel>` | Leave a channel. |
| `slack post [channel] [text]` | Post a message (plain text or Block Kit via `--blocks`). |
| `slack post-ephemeral <channel> <userId> <text>` | Post a message only one user in the channel can see. |
| `slack update <channel> <ts> <text>` | Edit a message the bot already posted. Pass `--blocks` to re-render it in place (e.g. update an interactive form or card); the text then becomes the notification fallback. |
| `slack delete <channel> <ts>` | Delete a message the bot posted. |
| `slack schedule <channel> <text>` | Schedule a message for future delivery. |
| `slack dm <userId> <text>` | Open a DM and send. One user ID for 1:1, or `U111,U222` for a group DM. |
| `slack history <channel>` | Top-level messages (paginated). Thread replies are marked; read with `slack thread`. |
| `slack threading` | Show or set the reply-in-thread preference for DMs and group DMs. |
| `slack thread <channel> <ts>` | Fetch replies in a thread. |
| `slack react` / `slack unreact` | Add or remove a reaction. |
| `slack reactions <channel> <ts>` | List the reactions on a message. |
| `slack upload [channel] [files..]` | Upload files to a channel or thread. |
| `slack file-download <fileId>` | Download a file shared in Slack into the current directory. |
| `slack users` / `slack user <id>` | Look up workspace users. |
| `slack channel-info` / `slack permalink` | Channel and message details. |
| `slack canvas <subcommand>` | Canvases: `list`, `read`, `create`, `edit`, `delete`, `lookup`, `set-access`, `remove-access`. |
| `slack bookmark <subcommand>` | Channel header bookmarks: `add`, `edit`, `remove`, `list`. |
| `slack pin <subcommand>` | Pin messages: `add`, `remove`, `list`. |
| `slack webhook <subcommand>` | Forward verified Slack events to your own webserver at `/webhook/slack`: `enable` (returns the delivery URL and a signing secret, shown once) and `disable`. |
| `slack mute` | Stay silent in the current channel or thread (not DMs) until someone @-mentions the agent again. |

### `slack post` flags

| Flag | What it does |
| - | - |
| `--thread <ts>` | Reply in that thread. |
| `--no-thread` | Post to the channel root instead of the current thread. |
| `--blocks` | Block Kit JSON string or `@path/to/file.json`. |
| `--force` | Allow posting into the conversation the agent is already answering (normally blocked to avoid duplicate replies). |
| `--broadcast` | When replying in a thread, also send to the channel. |

If channel is omitted, the current conversation is used where that makes sense.

## Reply threading in DMs

`platform slack threading` controls whether the agent replies inside a per-message thread in Slack DMs. It has two independent switches: `--dms` for 1:1 DMs and `--group-dms` for group DMs. Run it with no flags to print the current preference. Channels always thread and are not affected.

```bash theme={null}
# See the current preference
platform slack threading

# Thread replies in 1:1 DMs, keep group DMs flat
platform slack threading --dms on --group-dms off
```

The change takes effect from the next top-level message. See [Slack](/docs/channels/slack) for the same setting in the agent's profile.

## Example

```bash theme={null}
platform slack post C01234567 "Deploy finished. Changelog is up."
```

```bash theme={null}
platform slack post C01234567 "Fix is in the thread." --thread 1717000000.123456
```

```bash theme={null}
platform slack history C01234567 --limit 20
platform slack thread C01234567 1717000000.123456
```

```bash theme={null}
platform slack mute
```

See [Slack](/docs/channels/slack).


# todo
Source: https://getskydive.com/docs/api-reference/cli/todo

The live task list

`platform todo` maintains the task list that renders in the conversation as the agent works. Each `todo write` **replaces the full list** (not a patch). Mark exactly one item `in_progress` at a time.

| Subcommand | What it does |
| - | - |
| `todo write [items]` | Replace the list. JSON array, or `--file`, or stdin. |
| `todo list` | Print the current list. |
| `todo dismiss` | Hide the list when it is no longer relevant. |

Item shape: `{ "content": "<short task>", "status": "pending" | "in_progress" | "completed" }`.

## Example

```bash theme={null}
platform todo write "[
  {\"content\":\"Pull repo metadata\",\"status\":\"completed\"},
  {\"content\":\"Summarize open PRs\",\"status\":\"in_progress\"},
  {\"content\":\"Draft standup post\",\"status\":\"pending\"}
]"
```

```bash theme={null}
platform todo list
platform todo dismiss
```

See [Todos](/docs/get-started/talking-to-your-agent#todos).


# training
Source: https://getskydive.com/docs/api-reference/cli/training

Review work before it acts

`platform training` puts the agent's work in front of a person before it goes out. While [training mode](/docs/agents/training) is on, the agent presents each result as an artifact the builder reviews, rather than acting directly.

| Subcommand | What it does |
| - | - |
| `training activate` | Turn training mode on for the current thread. |
| `training deactivate` | Turn it off again. |
| `training create <kind>` | Create an artifact for the builder to review. |
| `training update` | Add a new revision to an existing artifact. |
| `training history` | List past artifacts and the feedback on them. |

The usual loop: activate, create an artifact, read the feedback, `update` with a revision, repeat until it is approved.

## Example

```bash theme={null}
platform training activate
platform training create draft
platform training history
```

See [Training mode](/docs/agents/training).


# Utility commands
Source: https://getskydive.com/docs/api-reference/cli/utility

Runtime info and maintenance

Maintenance and introspection commands the agent uses to manage its own runtime. Most of these relate to the agent's sandbox lifecycle; for the source of truth on what the agent *is* (its persona, tools, MCP config, and memory), see the [agent repository](/docs/agents/repo).

| Command | What it does |
| - | - |
| `info` | Print runtime environment details (tools, versions, capabilities). |
| `policies list` / `policies show <id>` | View policies assigned to this agent. |
| `trace list` / `trace query <id>` | Inspect the agent's own execution traces. |
| `logs` | Read the agent's sandbox logs. See below. |
| `restart` | Push work and restart the sandbox, picking up pending updates. |
| `harness reload` | Restart the harness for new tools or dependencies. Current response ends. |
| `update` | Check for and apply system updates immediately. |
| `self-upgrade` | Upgrade the runtime to the latest platform version, keeping agent-changed files. |

Template rebuild status lives under [`platform builds`](/docs/api-reference/cli/builds).

## `platform logs`

Read the agent's own sandbox logs: harness output, system logs, and proxy logs, oldest-first.

| Flag | What it does |
| - | - |
| `--since <time>` | Start from an absolute ISO timestamp or a relative duration (`30m`, `2h`, `1d`). |
| `--level <level>` | Filter by log level. |
| `--limit <n>` | Maximum number of entries to return (default 200, max 1000). |
| `--json` | Output as newline-delimited JSON. |

The agent can only read its own logs. Logs are scoped to the running sandbox.

```bash theme={null}
platform logs --since 30m
platform logs --level error --json
```

## `platform self-upgrade`

Upgrades the harness to the latest platform template version. The upgrade applies only upstream changes, so your edits to `webserver.sh`, `Dockerfile`, `package.json`, and other template-managed files are not touched.

```bash theme={null}
platform self-upgrade          # apply the upgrade
platform self-upgrade --dry-run  # preview what would change
platform self-upgrade --force    # override if the safe patch can't be computed
```

If the patch can't be computed from your recorded base version (for example, on a fresh clone with no upgrade history), `self-upgrade` will refuse unless you pass `--force`. `--force` generates a full content-reset patch, which can overwrite your custom template files. Use `--dry-run` first to see what it would affect.

## Example

```bash theme={null}
platform info
platform policies list
platform trace list
```


# web
Source: https://getskydive.com/docs/api-reference/cli/web

Web chat tools

`platform web` works with the web chat conversation: attachments, history, bringing other agents in, and archiving conversations.

| Subcommand | What it does |
| - | - |
| `web self` | Show this agent's web identity. |
| `web conversations` | List past web conversations. |
| `web history [conversationId]` | Read messages from a web conversation. |
| `web attachments` | List, show, download, and share files in web conversations. |
| `web add-agent <agent-id>` | Add another agent to the current conversation. Pass `--message "..."` to dispatch that agent's first turn immediately after it joins. |
| `web remove-agent [agent-id]` | Remove an agent. Omit the id to leave yourself. |
| `web archive <conversationId>` | Archive a conversation for the person you are acting for. |
| `web unarchive <conversationId>` | Restore an archived conversation to the default list. |
| `web create` | Start a new web conversation with an opening message. Alias for `platform conversations create`. |
| `web search <query>` | Search across this agent's past conversations. Alias for `platform conversations search`. |

Adding and removing agents only works in web conversations. A conversation holds at most three agents at once.

When you pass `--message`, the added agent receives a run dispatched for it as soon as it joins. Without `--message`, the join is silent and the agent waits until it is addressed:

```bash theme={null}
# Silent join
platform web add-agent 019e....

# Join and immediately give the agent its first turn
platform web add-agent 019e.... --message "You've been added to help with the data pipeline review. Start by reading the attached spec."
```

`--message` also works if the agent is already in the conversation (dispatch-only, no duplicate join).

## Archiving conversations

`platform web archive` and `platform web unarchive` let an agent tidy the conversation list for the person it is acting for. Archiving moves a conversation out of the default view for that user only. It does not delete anything, and it does not affect what anyone else sees.

```bash theme={null}
# Move a conversation out of the default list
platform web archive <conversationId>

# Bring it back
platform web unarchive <conversationId>
```

The conversation ID comes from `platform web conversations`. A 404 means the conversation does not exist, this agent is not a participant, or the resolved user is not in it.

> [!NOTE]
>
> Archiving is per-user and reversible. It is not a delete. The conversation stays in the system and other participants still see it normally.

## Starting and searching conversations

`web create` and `web search` are discoverable aliases for the corresponding [`platform conversations`](/docs/api-reference/cli/conversations) commands. The underlying behavior is identical; which command you reach for is up to you.

```bash theme={null}
# Start a new conversation with an opening message
platform web create --message "Weekly report is ready."
platform web create --message "Found 3 failing tests." --title "Test failures: 2026-08-06"

# Search across past conversations
platform web search "billing question"
platform web search "deploy script" --limit 5
```

For cross-channel list, full-transcript dumps, or JSONL output, use `platform conversations list` and `platform conversations show` directly.

## Example

```bash theme={null}
platform web self
platform web history
platform web add-agent 019e....
platform web conversations
platform web archive 019f....
platform web create --message "Starting the weekly audit now."
platform web search "auth setup"
```

See [Web chat](/docs/channels/web).


# workspace
Source: https://getskydive.com/docs/api-reference/cli/workspace

Invite members, manage invitations, and create teams

`platform workspace` and `platform invites` manage who belongs to the [workspace](/docs/workspaces/overview): sending member invitations, listing the ones still pending, and revoking them. `platform team` creates teams inside that workspace and adds or removes members.

| Subcommand | What it does |
| - | - |
| `workspace invite <emails..>` | Invite one or more people to the workspace by email. Optional `--role` (`member`, `admin`, or `owner`; privileged roles require an owner). |
| `invites list` | List pending member invitations. |
| `invites add <email> [role]` | Invite a member to the workspace (same as `workspace invite` for one address). |
| `invites cancel <invitation-id>` | Revoke a pending invitation. |
| `team create <name>` | Create a team in the workspace. |
| `team add-member <team-id> <user-id>` | Add a workspace member to a team. |
| `team remove-member <team-id> <user-id>` | Remove a member from a team. |
| `team delete <team-id>` | Delete a team. Fails if any agents still have access to it. |

The agent can only act within its own permissions: granting `admin` or `owner` requires the workspace owner. Creating and changing teams requires a workspace admin or owner.

## Example

```bash theme={null}
platform workspace invite alex@example.com sam@example.com --role member
```

```bash theme={null}
platform invites list --json
```

```bash theme={null}
platform invites cancel inv_01J9Z2
```

```bash theme={null}
platform team create Growth
```

See [Teams](/docs/workspaces/teams) for what members can do once they join, and [Roles & access](/docs/workspaces/rbac) for what each role allows.


# API overview
Source: https://getskydive.com/docs/api-reference/overview

Manage agents, API keys, and secrets over HTTP

The Skydive API is a REST API for managing agents and their credentials programmatically. It is the surface the Skydive CLI is built on: list and create agents, mint and revoke API keys, and manage the secrets an agent's sandbox can see.

```bash theme={null}
curl https://api.skydive.com/v1/agents \
  -H "Authorization: Bearer $SKYDIVE_API_KEY"
```

```json theme={null}
{
  "agents": [
    {
      "id": "019f6240-72f2-7a7d-acb0-ca83fc1072dc",
      "name": "Nephew",
      "url": "https://019f6240-72f2-7a7d-acb0-ca83fc1072dc.skydive.app",
      "description": "Engineering assistant",
      "model": "anthropic/claude-opus-4-8",
      "gitUrl": "https://git.skydive.storage/anyone-agent-019f6240-72f2-7a7d-acb0-ca83fc1072dc.git"
    }
  ],
  "count": 1,
  "nextCursor": null
}
```

All endpoints are served from a single versioned base URL:

```
https://api.skydive.com/v1
```

Requests and responses are JSON. The API is also self-describing: the current [OpenAPI document](https://api.skydive.com/v1/openapi.json) and an [interactive reference](https://api.skydive.com/v1/docs) are served by the API itself.

## Authentication

Every request requires an API key in the `Authorization` header:

```
Authorization: Bearer sky_live_...
```

Keys are created from an agent's **API access** settings in the app, or through the API itself. The full key value is returned exactly once, at creation. Skydive stores only a hash, so a lost key cannot be recovered, only replaced. Requests without a valid key return `401`.

> [!WARNING]
>
> An API key carries the authority to act on your agents. Keep it out of client-side code and version control. If a key is exposed, revoke it and mint a new one; revocation takes effect immediately.

### Scopes and binding

Every key has a scope and a binding, set at creation. Together they determine exactly what a request may do:

| Property | Values | Effect |
| - | - | - |
| Scope | `use` | Read operations: get and list agents, list key metadata, list secret names. |
| | `edit` | Everything `use` allows, plus create agents, mint and revoke keys, and set or delete secrets. |
| Binding | agent-bound | The key acts on one agent only. It cannot read or modify any other agent, and cannot list or create agents across the workspace. |
| | account-level | The key acts on any agent its owner can edit, and can list and create agents. |

A request that exceeds the key's scope or binding fails with `403` and an explanatory message. It never partially succeeds.

Keys minted through `POST /v1/agents/{agentId}/api-keys` are always agent-bound with the `use` scope.

## Errors

The API uses conventional HTTP status codes. Every error response has the same shape: an `error` field with a human-readable message, and on some errors a machine-readable `code`.

```json theme={null}
{
  "error": "this api key lacks the required 'edit' scope"
}
```

| Status | Meaning |
| - | - |
| `400` | The request is malformed: an invalid pagination cursor, or a secret name that is not a valid identifier. |
| `401` | The API key is missing, malformed, or revoked. |
| `403` | The key is valid, but its scope or agent binding does not permit this operation. |
| `404` | The resource does not exist or is not visible to you. Skydive returns `404` rather than `403` for agents you cannot view, so the existence of a private agent is never revealed. |
| `409` | The request conflicts with existing state, such as a name that cannot be assigned. |
| `429` | Agent creation quota or rate limit exceeded. |
| `5xx` | Something failed on Skydive's side. Retry with backoff. |

## Pagination

List endpoints return results newest first and paginate with an opaque cursor. Pass `limit` (1 to 100, default 50) to size the page, and pass the previous response's `nextCursor` to fetch the next page. A `null` `nextCursor` means you have reached the end.

```bash theme={null}
curl "https://api.skydive.com/v1/agents?limit=2" \
  -H "Authorization: Bearer $SKYDIVE_API_KEY"
# ... "nextCursor": "2026-08-01T09:14:52.000Z|019f62..."

curl "https://api.skydive.com/v1/agents?limit=2&cursor=2026-08-01T09%3A14%3A52.000Z%7C019f62..." \
  -H "Authorization: Bearer $SKYDIVE_API_KEY"
```

Treat cursors as opaque strings: URL-encode them when passing them back, and do not parse or construct them.

## Versioning

The path prefix is the API version. Within `v1`, response shapes are additive only: new fields may appear, but existing fields are never removed, renamed, or retyped. Breaking changes ship as a new version prefix. Write clients that ignore unknown fields.

## Secrets are write-only

Secret values can be set and deleted but never read back. `GET /v1/agents/{agentId}/secrets` returns key names only. This matches how secrets work inside the agent's sandbox and keeps the API from becoming a way to exfiltrate credentials. To rotate a secret, set the new value; to verify one exists, list the names.

## Endpoints

> **[Agents](/docs/api-reference/agents/list-agents)**
>
> List, fetch, and create agents.

> **[API keys](/docs/api-reference/api-keys/list-an-agents-api-keys)**
>
> Mint, list, and revoke an agent's keys.

> **[Secrets](/docs/api-reference/secrets/list-an-agents-secret-names)**
>
> Set and remove an agent's secrets.


# Platform CLI
Source: https://getskydive.com/docs/api-reference/platform-cli

Every command your agent can run to drive the platform

Every Skydive agent has a `platform` command-line tool inside its sandbox. It's how the agent drives the platform itself: connecting services, scheduling runs, messaging across channels, managing other agents, and more.

You typically won't run these yourself. You ask the agent in plain language and it picks the right command. This reference lists the full command surface so you can see exactly what your agent can do under the hood.

> [!NOTE]
>
> This is a deep reference, kept out of the main sidebar on purpose. Everything here is linked, indexed, and stable, so bookmarks and search results keep working. If you are looking for how to *use* an agent rather than what its runtime can do, start at [Agents](/docs/agents/overview).

> [!NOTE]
>
> This reference mirrors the agent's live CLI. The agent can always run `platform --help`, or `platform <command> --help`, for the current flags and options in its sandbox. Prefer `--help` when a detail might have changed.

> [!NOTE]
>
> Looking for the agent's config files, MCP connections, memory, or persona? Those live in the agent's [git repository](/docs/agents/repo), not in the CLI. The CLI is for runtime commands; the repository is the source of truth for everything the agent is.

## How to read this reference

* Each command page has a short intro, a subcommand table, and at least one **Example** you can copy.
* Prefer `platform <cmd> --json` when the agent needs structured output.
* Channel tools (`slack`, `email`, `web`, `imessage`) default to the **current conversation** when you omit the channel or thread.

## Command groups

> **[agent](/docs/api-reference/cli/agent)**
>
> Create and manage other agents.

> **[auth](/docs/api-reference/cli/auth)**
>
> Connect external services.

> **[secrets](/docs/api-reference/cli/secrets)**
>
> Store API keys securely.

> **[automation](/docs/api-reference/cli/automation)**
>
> Run on a schedule.

> **[db](/docs/api-reference/cli/db)**
>
> Provision Postgres databases.

> **[compute](/docs/api-reference/cli/compute)**
>
> Resize the sandbox.

> **[builds](/docs/api-reference/cli/builds)**
>
> Sandbox template build status.

> **[ask](/docs/api-reference/cli/ask)**
>
> Ask the user a multiple-choice question.

> **[todo](/docs/api-reference/cli/todo)**
>
> The live task list.

> **[feed](/docs/api-reference/cli/feed)**
>
> Post cards to the feed.

> **[notify](/docs/api-reference/cli/notify)**
>
> In-app bell, or omit --kind for the daily digest.

> **[training](/docs/api-reference/cli/training)**
>
> Review work before it acts.

> **[models](/docs/api-reference/cli/models)**
>
> List the models you can run on.

> **[deep-learn](/docs/api-reference/cli/deep-learn)**
>
> Background learning sessions.

> **[search](/docs/api-reference/cli/search)**
>
> Search and read the web.

> **[computer](/docs/api-reference/cli/computer)**
>
> Browser and file share links.

> **[portal](/docs/api-reference/cli/portal)**
>
> Reach your own computer.

> **[post](/docs/api-reference/cli/post)**
>
> Send a message to Slack, email, or iMessage.

> **[web](/docs/api-reference/cli/web)**
>
> Web chat tools.

> **[slack](/docs/api-reference/cli/slack)**
>
> Slack channel tools.

> **[email](/docs/api-reference/cli/email)**
>
> Email channel tools.

> **[imessage](/docs/api-reference/cli/imessage)**
>
> SMS and iMessage tools.

> **[channel](/docs/api-reference/cli/channel)**
>
> Channel context.

> **[conversations](/docs/api-reference/cli/conversations)**
>
> Past conversations.

> **[workspace](/docs/api-reference/cli/workspace)**
>
> Invite members, manage invitations, and create teams.

> **[billing](/docs/api-reference/cli/billing)**
>
> Check workspace billing state and resolve blocks.

> **[feedback](/docs/api-reference/cli/feedback)**
>
> Report platform problems.

> **[Utility commands](/docs/api-reference/cli/utility)**
>
> Runtime info and maintenance.


# Delete an agent secret
Source: https://getskydive.com/docs/api-reference/secrets/delete-an-agent-secret

/api-reference/openapi.json delete /agents/{agentId}/secrets/{key}
Removes one secret from the agent's vault. Succeeds with `204` whether or not the key existed, so deletes are safe to retry. Requires edit access to the agent and a key with the `edit` scope.



# List an agent's secret names
Source: https://getskydive.com/docs/api-reference/secrets/list-an-agents-secret-names

/api-reference/openapi.json get /agents/{agentId}/secrets
Returns the names of the secrets visible to an agent, including any inherited from its workspace. Values are never returned: secrets are write-only over the API. Use this to check whether a key exists before setting or rotating it.



# Set an agent secret
Source: https://getskydive.com/docs/api-reference/secrets/set-an-agent-secret

/api-reference/openapi.json put /agents/{agentId}/secrets/{key}
Creates or replaces one secret in the agent's vault. The value becomes available to the agent's sandbox as an environment variable and to any outbound proxy rule that references the key; a changed value takes effect on the agent's next request. Values cannot be read back. Key names must be env-var-style identifiers, for example `MY_API_KEY`. Requires edit access to the agent and a key with the `edit` scope.



# Browser
Source: https://getskydive.com/docs/capabilities/browser

Your agent drives a real web browser

Every agent has a real web browser it can drive: navigating pages, clicking, filling forms, and pulling data. Use it when a task can only be done inside a website.

![Cove using a browser](https://mintcdn.com/skydive/2VR7bKqFZZqW4hyR/images/agents/browser-cove.png?fit=max\&auto=format\&n=2VR7bKqFZZqW4hyR\&q=85\&s=bb558b993904eb8a1918552d8f187b6b)

## What it can do

> **Navigate & read**
>
> Open pages, follow links, and extract content and data.

> **Interact**
>
> Click buttons, fill forms, select options, submit.

> **Log in**
>
> Sign into sites, with your help for credentials when needed.

> **Capture**
>
> Take screenshots to show you what it sees.

## The shared browser

In [web chat](/docs/channels/web), the agent's browser is **shared with you**. It appears in the Computer panel. Watch it work, or take over whenever a human is faster:

* **Logins**: sign in yourself instead of handing over credentials.
* **Captchas & 2FA**: solve the human-verification step the agent can't.
* **Payments & confirmations**: take over for anything sensitive.

You drive the same browser the agent does. Use Back, Forward, and Reload. Type a URL or a search into the address bar. Open or close tabs. A tab you open is the agent's tab too: it shows up in the agent's tab list, so you can line up a page and hand the task back. The agent picks up where you left it.

On [Slack](/docs/channels/slack), [email](/docs/channels/email), or [text](/docs/channels/text), the same browser is still yours to reach. When a step needs your own login or a code, the agent names that step and links you to the web conversation. Open the shared browser there and enter it yourself. A credential that has to be yours is never a dead end.

## Terminal

The Computer panel also has a **Terminal** tab. Open it to get a live shell into the agent's sandbox, useful for inspecting what the agent built, running a quick command, or debugging something without asking the agent to do it for you. The terminal is only shown when you have edit access on the agent.

## Search vs. browse

The agent uses the lighter tool first. For reading public information it uses [web search and fetch](/docs/capabilities/web-search). It opens the full browser when it needs to **interact**: clicking, filling, logging in, or when a page needs a real browser to render.

## Good uses

* Pulling data from a dashboard that has no API
* Filling out a web form on your behalf
* Checking a page that requires a login
* Capturing a screenshot of how something looks
* Researching across sites that block simple fetching

Ask for it in plain language:

```text theme={null}
Log into our analytics dashboard and pull last week's signup numbers
into a spreadsheet, broken down by day.
```

> [!NOTE]
>
> For anything requiring authentication, a captcha, or a payment, the agent hands off to you in the shared browser.

> **[Connect accounts for repeat access](/docs/integrations/oauth)**
>
> Authorize a service once so the agent doesn't need the browser to log in every time.


# Custom domains
Source: https://getskydive.com/docs/capabilities/custom-domains

Serve your agent's website from your own domain instead of skydive.app

Every agent's website gets an address like `myagent.skydive.app`. If you want it on a domain you own, say `app.yourcompany.com`, you can connect one in a couple of minutes.

> [!NOTE]
>
> Custom domains are rolling out to workspaces gradually. If you don't see the option yet, it hasn't reached your workspace.

## Where to set it up

Open your agent's profile, find the **Connections** card, and open the site address settings (the "..." menu on the Website URL row). Choose **Set up a custom domain**.

An agent has one site address at a time. When a custom domain is live, it replaces the `skydive.app` label; remove the custom domain and the agent falls back to its `skydive.app` address automatically.

## Setting up the domain

Enter your host (for example `app.yourcompany.com`) and choose how to add the DNS records:

**Set up automatically** is the default. You sign in to your DNS provider (Namecheap, GoDaddy, and others) through a secure dialog, and the records get written for you. When it finishes, Skydive verifies the domain on its own. If automatic setup is unavailable, you'll see a message saying so and can use the manual path instead.

**Set up manually** shows you the records to publish at your registrar:

* A `TXT` record that proves you own the zone.
* A `CNAME` that sends visitors to your agent. At an apex domain (like `yourcompany.com`), what Skydive asks for depends on your DNS provider: where the provider supports CNAME flattening, you publish a single `CNAME` at the root; where it doesn't, you get `A` records instead. A `www` CNAME is included either way so `www.yourcompany.com` serves the site instead of your registrar's parking page.

Click **Verify** after the records are published.

## While DNS propagates

After the records are in place, the modal re-checks every 15 seconds while it's open. Each record row reports its own status, so you can see which records DNS has picked up and which are still missing. The overall banner keeps waiting until the hostname is fully active (which can trail the individual records), then verifies the TLS certificate. If you close the modal or reload the page mid-setup, your progress is saved and the next check picks the flow back up.

DNS changes can take from a few minutes to a few hours to appear, depending on your provider. If verification fails right after publishing records, give it a few minutes and verify again.

## Troubleshooting

* **"Automatic setup is unavailable right now"**: the automatic flow couldn't be reached. Use **Set up manually** and publish the records yourself.
* **Verification stuck on the TXT record**: some registrars cache DNS misses for up to an hour, so a freshly published record can stay invisible to verification for a while. Wait and verify again.
* **`www.yourdomain.com` shows an old page**: the `www` CNAME wasn't published or still points at the registrar's default. Update it to the same value as the main record.
* **A green record in your registrar but the modal still shows it missing**: re-run the setup and it re-checks everything, including records that were missing from the first pass.
* **Every record row shows valid but the banner still says waiting**: the row statuses read live DNS, while the overall check waits until the hostname is active at the serving layer, which can lag the records by a bit. Give it a few minutes.

For anything else, email [support@skydive.com](mailto:support@skydive.com) with your domain and a screenshot of the setup modal.


# Files
Source: https://getskydive.com/docs/capabilities/files

A shared folder for you and your agent

Every agent has a **shared workspace**: a folder that both you and the agent read and write. Deliverables show up there. You drop files there for the agent. Work-in-progress lives there between sessions.

## The shared workspace

In [web chat](/docs/channels/web), the workspace appears as a **Files** tab you can browse, open, and download from. Both sides have access:

> **The agent saves here**
>
> Reports, exports, generated images, and anything it produces for you show up automatically.

> **You drop files here**
>
> Add reference material or inputs for the agent to use.

The workspace is organized like a folder someone else will browse, with clear names and subfolders. It **persists across sessions**: what's saved is still there next time.

Open a file to preview it in the panel. Documents, images, spreadsheets, and CSVs render inline. Markdown files (`.md`) show formatted, with a toggle for the raw source.

## Attachments

You can also send files directly in any conversation: on [web chat](/docs/channels/web), [Slack](/docs/channels/slack), [email](/docs/channels/email), or [text](/docs/channels/text). The agent fetches each attachment, works with it, and returns results.

Agents handle a wide range of files:

* **Documents**: read, summarize, extract, rewrite
* **Spreadsheets & data**: analyze, transform, chart
* **Images**: describe, edit, convert
* **Audio**: transcribe to text
* **PDFs**: pull out content and data

For example, drop in a signed contract and ask:

```text theme={null}
Pull the key dates and renewal terms out of this contract into a table.
```

## Getting results back

When an agent produces something for you to keep, it saves it to the shared workspace, where it persists and you can browse it. It can also attach it as a downloadable file in the conversation. Either way, deliverables come to you.

> [!TIP]
>
> Need the agent to keep a file long-term? Ask it to save to the workspace. Scratch and intermediate work it keeps elsewhere, so the shared space stays clean and browsable.

## Browsing files on the agent's profile

The agent's profile page has a **Files** tab that shows everything in its shared workspace, across all conversations. Images appear as a thumbnail grid you can open in a lightbox. Documents appear as a list with a download button on hover. Use the filter to switch between All, Images, and Documents.

Files here are the same ones the agent saves during its work. There's no delete from the profile view; to remove a file, ask the agent to delete it.

> **[Build a live app instead](/docs/capabilities/webserver)**
>
> When you want something interactive rather than a file.


# Network controls
Source: https://getskydive.com/docs/capabilities/network

What the agent's computer can reach, and how credentials cross the wire

Every request leaving an agent's computer goes through Skydive's transparent proxy. Nothing reaches the internet directly. That single chokepoint is what makes the credential model work and what gives your workspace a place to govern outbound access.

## The proxy holds the credentials

When an agent calls a connected service, it sends the request without a token. The proxy matches the request against the rules for that agent and attaches the credential on the wire.

The agent never sees the secret. It cannot print it, log it, paste it into a conversation, or leak it through a prompt, because it was never in the sandbox to begin with.

> **[How this works in detail](/docs/integrations/permissions)**
>
> Scoping, revocation, and why the design is shaped this way.

## Rules are matched per agent

Proxy rules match on hostname, path, and method, and they are attached per agent. An agent gets a credential only for the rules it has been granted, so connecting GitHub for one agent does not hand GitHub to every agent in the workspace.

A workspace can override a rule for its own tenancy or substitute its own values, without changing anything for anyone else.

## Governing what leaves

Because every request crosses the same boundary, outbound access can be governed centrally rather than agent by agent. That is how workspace-level controls over which services agents may reach are enforced, and how usage is attributed back to the agent and workspace that caused it.

> [!NOTE]
>
> Control-plane traffic is never blocked: an agent that cannot reach a model can still report the problem, checkpoint its work, and tell you why it stopped. A cut-off agent that also goes silent is worse than a cut-off agent.

## What this means in practice

* **Secrets stay out of the sandbox.** Store them once, and the agent uses them without holding them.
* **Access is revocable in one place.** Remove a connection and the next request has no credential attached, with nothing to clean up inside the agent.
* **Reach is auditable.** Outbound calls are attributable to an agent and a workspace.

> **[Secrets](/docs/integrations/secrets)**
>
> Storing credentials the agent can use but never read.

> **[Enterprise controls](/docs/workspaces/enterprise)**
>
> SSO, audit logs, and workspace-level policy.


# Overview
Source: https://getskydive.com/docs/capabilities/overview

Every agent has its own computer in the cloud

Every Skydive agent has its own computer: an isolated Linux machine in the cloud, provisioned on demand and separate from every other agent's. Anything you can do on a computer, your agent can do there.

That is what separates an agent from a chatbot. It is not describing the work to you, it is doing the work on a machine.

## What is on it

> **[Browser](/docs/capabilities/browser)**
>
> A real browser it drives. Fill in forms, click through dashboards, pull what has no API.

> **[Files](/docs/capabilities/files)**
>
> A shared folder you both read and write. Deliverables land here.

> **[Terminal](/docs/capabilities/terminal)**
>
> A full shell. Run code, install packages, drive command-line tools.

> **[Search](/docs/capabilities/web-search)**
>
> Search the web and fetch pages, without opening a browser.

> **[Apps & webserver](/docs/capabilities/webserver)**
>
> Host a live web app at a URL you can visit.

> **[Storage](/docs/capabilities/storage)**
>
> A managed Postgres database for anything that has to persist.

## It keeps working when you stop

The computer is not your laptop, so nothing depends on your laptop staying awake. Close the lid and the agent keeps going: a long build finishes, a [routine](/docs/agents/routines) fires at 3am, a browser session works through a queue.

The sandbox sleeps when idle and wakes on demand. Work that must survive a restart goes in the agent's [repo](/docs/agents/repo), its [files](/docs/capabilities/files), or [storage](/docs/capabilities/storage).

## Two computers, one job

Some work lives on your machine instead: a repo you have checked out, a dev server on `localhost`, a tool only installed locally. [Portal](/docs/integrations/portal) is the bridge. The agent reads what it needs from your machine, then moves the work up to its own computer and carries on.

> **[Sizes](/docs/capabilities/sizes)**
>
> How much machine an agent gets, and how it asks for more.

> **[Network controls](/docs/capabilities/network)**
>
> What the computer can reach on the internet, and how that is governed.


# Sizes
Source: https://getskydive.com/docs/capabilities/sizes

How much machine an agent gets, and how it asks for more

Every agent's computer has a size: how much CPU and memory it gets. Most agents never need to think about it. Some jobs do.

## The ladder

Sizes run from smallest to largest:

```
xsmall  <  small  <  medium  <  large  <  xlarge  <  xxlarge  <  xxxlarge
```

An agent can see its own size and which sizes it could move up to. It cannot resize itself.

## Asking for more

When a job is genuinely compute-bound, the agent requests a bigger machine and explains why. A test suite that runs out of memory, heavy parallel work, or a large data pass are the usual reasons.

**1. The agent asks**

It requests a specific size with a reason, and an Approve or Deny card appears in the conversation.

**2. Someone decides**

A person who can edit that agent approves or denies it. Cost is a human's call, which is why the agent cannot do this on its own.

**3. It restarts at the new size**

On approval the size is raised and the sandbox restarts, so in-flight work is checkpointed first. The agent resumes once it is back.

An agent can only request a size above its current one, and only one request can be open at a time.

> [!NOTE]
>
> A resize request needs a live conversation, because it needs someone to approve it. A [routine](/docs/agents/routines) firing with nobody watching cannot request one.

## When a bigger machine is not the answer

Reach for a resize when the work does not fit in memory or takes far too long on the current size. Do not reach for it when the real problem is scope: an agent doing five unrelated jobs is better split into several agents, or given [subagents](/docs/capabilities/subagents) to spread the work.

> **[Track what it costs](/docs/account/usage)**
>
> Compute draws on your credit balance alongside model usage.


# Storage
Source: https://getskydive.com/docs/capabilities/storage

Your agent can provision a managed Postgres database for the apps it builds

When an agent builds something that needs to remember data between visits, it provisions a managed Postgres database as the backend. No setup on your end. The agent creates the database, connects to it, and keeps the connection details safe.

![Sage organizing database objects](https://mintcdn.com/skydive/2VR7bKqFZZqW4hyR/images/agents/databases-sage.png?fit=max\&auto=format\&n=2VR7bKqFZZqW4hyR\&q=85\&s=12aabc4ccb994ddfc12470cd5dc34334)

## When an agent uses one

> **App backends**
>
> Persist the data behind a [web app](/docs/capabilities/webserver): records, settings, user accounts.

> **Dashboards**
>
> Store the rows a live dashboard reads from, so numbers survive a restart.

> **Accumulated data**
>
> Keep results from [routines](/docs/agents/routines) building up over time.

> **Structured storage**
>
> Anything that's cleaner as real tables than as files in the workspace.

## How it works

Ask the agent for something that needs to remember data. It provisions a database on its own:

```text theme={null}
Build me a small CRM where I can add contacts and notes, and keep
them between sessions.
```

The agent creates a Postgres database (hosted on Neon), attaches it under an environment variable, and wires its app to read and write from it. The connection string is stored as a secret in the agent's sandbox and never printed in chat.

> [!TIP]
>
> A database persists independently of the sandbox. The sandbox is wiped when it goes idle. The database and everything in it stays.

## Sizing

Databases come in three sizes: small, medium, and large. Most apps run well on the default small tier. The agent picks a larger size when the work needs more storage or throughput. You can also ask it to resize.

## Shared across your workspace

Databases belong to your workspace, not to a single agent. Any agent in the workspace can attach an existing database and connect to it. Two agents can share one backing store: one writes the data, another reads it. Each agent chooses the environment variable it attaches the database under. When a database is no longer needed, an agent can archive it. Archiving preserves the data and frees the name for reuse.

> **[Build the app that uses it](/docs/capabilities/webserver)**
>
> Databases are the backing store for the live web apps your agent hosts.


# Subagents
Source: https://getskydive.com/docs/capabilities/subagents

Agents delegate parallel work to isolated helpers

For a big job, an agent can split the work and hand parts of it to **subagents**: fresh, isolated copies of itself that each run on their own and report back. The agent stays in charge, gathers the results, and gives you one answer.

## Why it helps

A single agent works through a task one step at a time. Subagents let it work on several fronts at once.

* **Parallel work.** Three independent lookups run at the same time instead of one after another, so the whole job finishes sooner.
* **A clean workspace.** A large or noisy subtask (reading a long thread, crawling a site, sifting a data dump) runs in the subagent's own context, so it doesn't crowd out what the main agent is holding in mind.
* **A focused helper.** A subagent can take on an extra persona for its task, so one part of the work runs with a specialist's framing while the rest stays general.

Each subagent is the same agent, with the same memory, connected services, and identity. It just runs on its own for one scoped task.

## What it looks like

When an agent delegates, a card appears showing the work in flight. This works in both web chat and the `skydive chat` CLI. Its header reads **Subagent working** for a single helper, or **Subagents working in parallel** for a fan-out. Each subagent is one row, named for its task.

```text theme={null}
Subagents working in parallel
  ◐ Pull Q3 numbers from analytics
  ◐ Draft the customer recap
  ◐ Check the contract for renewal terms
```

In web chat, each row links to that subagent's conversation: Cmd/Ctrl-click or middle-click a row to open the child thread and watch the work. You can also expand the card or dismiss it. In the CLI, the card streams live in the transcript. Each row shows the task name, a live activity line, and an elapsed time and context-token count as the subagent works. Use `↑` / `↓` to select a row and press `↵` to open that subagent's conversation directly. `esc` returns you to the parent chat. Fan-outs wider than eight rows collapse into a `↓ N more` tail while running. As each subagent finishes, its result flows back to the main agent. Once the batch settles, the transcript keeps a short **Subagents finished running** divider where the work happened, and the rows remain navigable so you can review what each helper did.

## When agents use it

The agent reaches for subagents when the shape of the work fits, not on every task:

* Several independent pieces that can run at once (research across three sources, checks against multiple systems).
* A heavy subtask worth keeping out of the main thread (a long document to read, a large result set to distill).
* A part of the job that benefits from a specialist framing.

For a quick, single-threaded task, the agent just does it inline. No card appears.

## Cost

Each subagent is a full agent run and counts toward usage like any other run. A well-scoped subtask (a search, a summary, a mechanical edit) often runs on a lighter, cheaper model, so a fan-out of small jobs stays inexpensive. See [Usage](/docs/account/usage) for how runs are metered.

> **[See work across conversations](/docs/workspaces/feed)**
>
> The feed shows what your agents are doing everywhere, not just the current chat.


# Terminal
Source: https://getskydive.com/docs/capabilities/terminal

A full shell on the agent's computer

The agent's computer runs Linux, and the agent has a full shell on it. This is how it runs code, installs what it needs, and drives command-line tools.

You never have to ask for this in shell terms. Describe the outcome and the agent picks the commands.

## What it uses the shell for

* **Running code.** Python, Node, whatever the job needs, on real data rather than in a description of the work.
* **Installing tools.** It adds packages when a task calls for them.
* **Processing files.** Converting, extracting, transforming, and compressing what is in [files](/docs/capabilities/files).
* **Driving CLIs.** Any tool that ships a command-line interface, including `git` and `gh`.
* **Long jobs.** Builds and batch work run in the background while the conversation continues.

## Opening it yourself

The Computer panel has a **Terminal** tab: a live shell into the agent's sandbox. It is useful for inspecting what the agent built, running a quick command, or debugging without asking the agent to do it for you.

The terminal is only shown when you have edit access on that agent.

> [!TIP]
>
> You can also reach the sandbox from your own machine. In the [Skydive CLI](/docs/cli/overview), `/sandbox` opens an interactive terminal, and `/sandbox <command>` runs one command and streams the output.

## What persists

The sandbox is ephemeral. It sleeps when idle and is rebuilt when it wakes, so anything installed at runtime is gone on the next boot.

Work that has to survive goes in one of three places:

| Where | For |
| - | - |
| The agent's [repo](/docs/agents/repo) | Config, tools, memory, and anything that defines the agent |
| [Files](/docs/capabilities/files) | Documents and deliverables you share with the agent |
| [Storage](/docs/capabilities/storage) | Structured data an app reads and writes |

For a package the agent needs on every boot, it adds the install line to its own Dockerfile in the repo, so the next sandbox is built with it.

> **[Apps & webserver](/docs/capabilities/webserver)**
>
> Host what it builds at a URL you can open.

> **[Sizes](/docs/capabilities/sizes)**
>
> When a job needs a bigger machine.


# Tools & MCP
Source: https://getskydive.com/docs/capabilities/tools-and-mcp

What your agent can do, and how it gives itself more

Every Skydive agent starts with a built-in toolset. It isn't limited to that set. When a job needs something the agent doesn't have, it gives itself the capability and keeps going.

![Otis connecting tools and MCP servers](https://mintcdn.com/skydive/2VR7bKqFZZqW4hyR/images/agents/tools-mcp-otis.png?fit=max\&auto=format\&n=2VR7bKqFZZqW4hyR\&q=85\&s=e3d4544b4348121e8dd11f4bd5d23024)

## Built-in tools

Out of the box, every agent can:

> **Run code & commands**
>
> Execute shell commands and scripts in its sandbox.

> **Read & write files**
>
> Create, read, and edit files: documents, data, code.

> **[Browse the web](/docs/capabilities/browser)**
>
> Drive a real browser to navigate, click, and extract.

> **[Search the web](/docs/capabilities/web-search)**
>
> Find and fetch up-to-date information.

> **[Host web apps](/docs/capabilities/webserver)**
>
> Run a live web server you can visit.

> **Process media**
>
> Work with images, audio, PDFs, and more: transcribe, convert, transform.

## It can give itself new tools

The built-in set is the floor, not the ceiling. The agent has a real computer, so it can reach for whatever a task demands:

* **Install a command-line tool** and use it.
* **Call any API**, public or internal, once you've [connected the service](/docs/integrations/oauth) or given it a [key](/docs/integrations/secrets).
* **Write and run its own code** to compute, transform data, or automate a one-off job.
* **Add a new integration** for itself when a task needs a system it isn't connected to yet.

You don't pick from a fixed menu of features. You describe the outcome, and the agent assembles the means.

```text theme={null}
We don't have an integration for our internal billing API. Read its
docs, write a script to pull this month's overdue accounts, and post
the list to #finance.
```

No one set that up ahead of time. The agent reads the docs, writes the code, and uses the access you granted.

## Connecting to more systems with MCP

One way an agent extends itself is **MCP (Model Context Protocol)**: the open standard for connecting agents to external tools and data sources. If a capability exists as an MCP server, your agent can plug it in and use it.

### What MCP adds

* Tools that already exist as MCP servers (hosted or local)
* Connections to internal systems and data sources
* Provider-specific actions exposed through their own MCP servers

### Authenticated MCP servers

Many MCP servers require authorization. You authorize once. A proxy then attaches the credential to every future request automatically. Tokens never sit in plain config and never appear in the agent's prompts. See [Connecting services](/docs/integrations/oauth).

The Connect card for an MCP server grants access at whatever level that MCP app asks for (often the whole workspace or account). It is **not** the same as the OAuth connect flow some apps use, where you pick individual pages or records. On MCP Connect, you will not choose items one by one, even for apps (like Notion) that offer a picker on the regular OAuth path.

### Which tools from a server

By default the agent gets every tool the server advertises. In [`mcp.config.json`](/docs/agents/repo) you can set `tools` to keep only those names, or `excludeTools` to drop specific ones. If you set both, `tools` applies first.

## Choosing the right tool

The agent decides which tool fits each step. It searches when it needs facts. It browses when it needs to interact with a page. It runs code when it needs to compute or transform. It reaches for a new tool when nothing built-in fits. You describe the outcome; it picks the means.

> **[Browser use](/docs/capabilities/browser)**
>
> When the agent needs to act on a real web page.

> **[Connecting services](/docs/integrations/oauth)**
>
> Authorize the accounts and APIs your tools need.

> **[MCP config in the agent repository](/docs/agents/repo)**
>
> MCP connections and custom tools are files in the agent's git repository. Inspect or edit them directly.

> **[Platform CLI reference](/docs/api-reference/platform-cli)**
>
> Every command the agent can run inside its own sandbox.


# Search
Source: https://getskydive.com/docs/capabilities/web-search

Find and fetch up-to-date information

Agents search the web and read pages directly, without opening a full browser. Their answers and research draw on current information.

## Two ways to pull from the web

> **Search**
>
> Query the web and get back relevant results with their content inline.

> **Fetch**
>
> Given a URL, read the page's content directly.

These are the agent's first choice for finding information. They are quick and lightweight. The agent escalates to the full [browser](/docs/capabilities/browser) only when it needs to interact with a page.

## When agents search

* Answering questions that need current facts or data
* Researching companies, people, products, or topics
* Pulling documentation or reference material
* Checking something the agent isn't confident about, instead of guessing

A typical request:

```text theme={null}
Research our top 3 competitors' pricing and packaging, and summarize
the differences in a short table with links to each source.
```

## Grounded answers

The agent fetches real sources, so its answers reflect what is on the page right now, not what a model memorized. For research tasks, it gathers across multiple sources and synthesizes a result.

> [!TIP]
>
> When accuracy matters, ask the agent to cite its sources: "research X and include the links you used."

> **[Go deeper with a research session](/docs/agents/training#sending-it-away-to-study)**
>
> For big topics, start a background learning session.


# Apps & webserver
Source: https://getskydive.com/docs/capabilities/webserver

Your agent can build and host live web apps

Agents can host a live web server. Ask for dashboards, forms, landing pages, and interactive apps, and visit them in your browser.

![Linus deploying a web app](https://mintcdn.com/skydive/2VR7bKqFZZqW4hyR/images/agents/webserver-linus.png?fit=max\&auto=format\&n=2VR7bKqFZZqW4hyR\&q=85\&s=0ced31d29c3b17d3b9996ad42dd32e7b)

## What you can ask for

> **Dashboards**
>
> A live view of your data, metrics, or status.

> **Forms & tools**
>
> Internal tools and intake forms your team can use.

> **Landing pages**
>
> A page to share, complete with real interactivity.

> **Interactive apps**
>
> Anything you'd view and click through in a browser.

## How it works

When you ask an agent to build something you'd view in a browser, it stands up a web server in its sandbox and gives you a URL. The app is publicly accessible at that link. In [web chat](/docs/channels/web), you can view it inline and click through it as the agent iterates.

```text theme={null}
Build me a simple dashboard that shows our weekly signups as a chart,
with a date-range filter.
```

The agent builds it, hosts it, and shares the link. You review it live and ask for changes in the same conversation.

## It stays up

The agent's web server is managed for you. It keeps running and restarts if it crashes, so the app stays available at its URL. Ask the agent to change it and it redeploys in place.

## Iterating

Treat the first version as a draft and refine in passes:

```text theme={null}
Looks good. Make the chart bars our brand orange, add a total at the
top, and default the filter to the last 30 days.
```

> **[Share files instead of an app](/docs/capabilities/files)**
>
> For documents and exports rather than a live app, use the shared workspace.

> **[webserver.sh in the agent repository](/docs/agents/repo)**
>
> The web server startup script lives in the agent's git repository. Clone it to inspect or modify how it launches.

## Receiving webhooks

The same web server that hosts your apps can receive inbound HTTP events from external services. When Stripe processes a payment, GitHub merges a PR, or Intercom receives a new ticket, those services can POST an event to your agent's URL. Your agent wakes up, handles it, and goes back to sleep.

This is how you replace a polling routine with something leaner: instead of checking a service on a schedule, you let the service tell you when something happens.

```text theme={null}
Set up a webhook receiver for Intercom. When a new ticket arrives,
categorize it and post a summary to #support.
```

The agent sets up the receiver, gives you the endpoint URL, and you register it with the external service. From that point on, events arrive and get handled automatically.

> [!TIP]
>
> Every agent comes with a **webhook-receiver skill** pre-installed. It handles the full setup: fast acknowledgment so the provider doesn't time out, HMAC signature verification, and deduplication across runs. Ask the agent to build a webhook receiver and it loads the skill automatically.

### What the flow looks like

**1. Ask the agent to set up a receiver**

Tell it which service and what to do with each event. The agent creates the route on its web server and gives you the public endpoint URL.

**2. Register the URL with the external service**

Paste the URL into the service's webhook settings and choose the events to send. Most services send a validation ping immediately, so the agent needs to be listening before you save.

**3. Events arrive and get handled**

The agent wakes on each inbound POST, verifies the payload, and runs the action you described. No polling. No idle runs.

### Works with any provider

Any service that can POST to a URL works: Stripe, GitHub, Linear, Intercom, and others. The agent's web server is publicly accessible at a stable URL, so there is nothing special to configure on the Skydive side.

> **[Webhook-receiver skill](/docs/agents/skills)**
>
> Pre-installed on every agent. Covers signature verification, acknowledgment, and deduplication.

> **[Reduce costs with event-driven runs](/docs/account/cost-tips)**
>
> Why waking on a webhook is cheaper than polling on a schedule.


# Changelog
Source: https://getskydive.com/docs/changelog

New features, improvements, and fixes from the Skydive team.

## September 6th

* Added a guided onboarding experience that can research a user's professional context, suggest useful starting directions, and continue from web to iMessage.
* Added per-routine model selection and automation run history, making scheduled work easier to tune and audit.
* Added Mermaid diagram rendering in web chat and richer live-browser link previews.
* Added workspace-managed AI provider connections and stopped metering model costs when a workspace uses its own provider credentials.
* Improved Slack and iMessage with clearer connection prompts, more concise progress updates, durable question cards, and more reliable message delivery.

## August 30th

* Added per-conversation model selection so each chat can use the model best suited to the task.
* Added browser-based integration sign-in through agent chat, with clearer recovery when a sign-in link expires.
* Added an on-demand command to run a routine immediately and let agents manage their own routines from any conversation.
* Expanded Skydive Portal with machine-first access, piped terminal input, tunnel removal, and more reliable reconnects across sleep and network changes.
* Improved Slack with cleaner progress presentation, better thread handling, and safer attachment delivery.

## August 23rd

* Expanded the integrations catalog to more than 600 products, with clearer categories, logos, connection methods, and management controls.
* Added guided iMessage number setup and support for another messaging provider, improving availability for individual and group conversations.
* Added a canary release channel for Skydive Desktop while keeping stable releases intentionally versioned.
* Added temporary browser result screenshots and improved live-browser tabs, sign-in handoffs, mobile sharing, and recovery.
* Improved routines with clearer completed-run summaries and a Run now command for immediate execution.

## August 16th

* Added isolated browser sessions so agents can run independent browser tasks without interfering with one another.
* Added organization controls for zero-data-retention model selection, plus full-catalog search and model-listing tools for delegated work.
* Expanded Skydive Portal with durable machine identity, secret import, faster tunnels, and stronger ownership controls.
* Added inline agent media, cross-workspace shared links, and improved long-file previews.
* Improved Slack, iMessage, integrations, and subagents with more reliable delivery, setup, recovery, and error reporting.

## August 9th

* Added Portal file operations so agents can read, write, and edit files on an authorized local computer.
* Improved shared browser links so recipients see the full browser session rather than only the active tab.
* Added more reliable model fallback, browser recovery, and long-running agent execution.
* Improved the web and CLI experience with clearer status, navigation, and recovery states.

## August 2nd

* Improved Skydive Portal connectivity and machine presence checks for more reliable local-computer access.
* Added clearer Slack workspace context and safer channel handling for agents connected across multiple workspaces.
* Improved shared browser viewing and recovery when sessions reconnect.
* Improved agent and conversation reliability across long-running work, interrupted sessions, and reconnects.

## July 26th

* Added a substantially improved Skydive CLI: interactive workspace switching, conversation history and recaps, clearer agent descriptions, terminal notifications, local shell commands with `!`, and readable diffs for file edits.
* Added plan management and one-time credit purchases, with clearer checkout outcomes and notifications when a workspace is low on credits or a refill fails.
* Improved Slack with group-DM support, a clearer connect button, better link formatting, and a dedicated installation page.
* Added a mobile chat layout, Markdown file previews, image lightbox navigation, remembered Files view preferences, and confirmation before deleting a conversation.
* Improved account security with breached-password checks and failed-login rate limits.

## July 13th

* Introduced a step-by-step onboarding experience with clearer goals, integration setup, agent customization, and a more polished first conversation.
* Expanded the Skydive CLI with agent creation, model switching, progressive Markdown, image pasting, themes, prompt history, completion notifications, chat archiving, and actionable connection cards.
* Added Skydive Portal so an agent can securely work with a user-authorized local machine, plus temporary browser and file handoff links between channels.
* Improved Slack conversations with smoother single-message streaming, better identity resolution, optional per-message threads in DMs, and richer rendering of bot replies.
* Added persistent chat drafts, better file actions, links from feed posts back to their source chats, and a workspace-wide default model setting.
* Improved reliability for long-running routines, dropped browser connections, provider failures, and agent context limits.

## July 6th

* Launched the interactive `skydive chat` terminal experience, including workspace commands and non-interactive print mode for automation.
* Added scoped Google integrations for Drive, Gmail, Calendar, Docs, Sheets, Slides, and Forms, plus a user-facing way to disconnect OAuth connections.
* Added AI-generated routine suggestions and direct links from routine runs to their conversations.
* Added new model choices and automatic provider failover so agents can continue when a model provider is unavailable.
* Improved Slack with pinned messages and bookmarks, clearer connection states, deep links to connected channels, and a branded welcome experience.
* Added browser URL entry, agent avatar editing, workspace member and role management, template-build visibility, and clearer work-duration indicators.

## June 29th

* Moved Skydive to `skydive.com` and added first-party product documentation, blog, and service-status links.
* Added workspace-level model-training opt-out controls, agent privacy badges, and confirmation when privacy settings change.
* Added an immutable administrative audit log with human-readable actor names.
* Improved onboarding on mobile, made workspace selection persist across sessions, and made it easier to understand who is connected to each integration.
* Added whole-page attachment drop, better routine editing through chat, per-tool activity summaries, and improved feed and chat navigation.
* Published clearer terms and privacy language for Skydive.

## June 22nd

* Added agent privacy levels, with access enforced across conversations, channels, activity, and realtime updates.
* Added a Finder-style Files experience with grid/list views, fullscreen previews, downloads, and support for files received through Slack.
* Expanded integrations with self-serve OAuth connections, clearer connection health, and new Aha! and HubSpot options.
* Added persistent agent browsers, agent-requested compute upgrades, managed Postgres databases, and dedicated agent email inboxes.
* Added richer routines, in-app notifications, searchable agents/posts/routines, and public links for shared posts.
* Improved first-run onboarding, mobile layouts, Slack multi-agent conversations, browser continuity, and recovery from long conversations.
* Added first-party Terms of Service and Privacy Policy pages.

## June 15th

* Launched a new marketing homepage and a redesigned web experience for the feed, agents, chat, routines, files, and workspace settings.
* Added richer agent profiles, customizable avatars, templates, group chats, shared posts, notifications, and recent-conversation discovery.
* Added user file uploads and workspace files, plus clearer live activity while an agent works.
* Added configurable compute tiers and improved agent reliability during long conversations and reconnects.
* Introduced Skydive CLI and API-key access for programmatic use.
* Introduced Skydive Desktop and Portal foundations for connecting agents to a Mac, along with true agent-to-agent subtask support.
* Improved Slack with streamed replies, synchronized thread history, interactive action links, and better Markdown rendering.

## June 8th

* Introduced the new Skydive web interface with a redesigned feed, chat, sidebar, attachments, and faster navigation.
* Added live tool-call activity and smoother text streaming so it is easier to follow an agent's work.
* Added self-serve integrations, including native OAuth for remote MCP services and managed Slack app provisioning.
* Added a workspace-wide team directory and a workspace file browser.
* Brought the product online at `getskydive.com` as Skydive's first branded home.
* Improved Slack reply clarity, scheduled-run reliability, chat error reporting, and recovery from interrupted agent sessions.

## June 1st

* Added a model switcher and natural-language policies for controlling how individual agents work.
* Added CSV table previews, richer file rendering, and better preservation of conversation context when editing or branching messages.
* Added self-updating Skydive command-line components and automatic recovery for unhealthy agent runtimes.
* Let agents propose new channel connections and made agent-sent email activity easier to identify.
* Improved Slack formatting, uploads, threading, direct messages, and error reporting.
* Improved reliability for long conversations, navigation, reconnects, and interrupted runs.

## May 25th

* Added web search and a persistent Documents folder for each agent.
* Added a full file browser with HTML and video previews, word wrapping, drag-and-drop attachments, and recent-file tabs.
* Added conversation forking, message steering, deeper learning workflows, agent duplication, and a usage page.
* Expanded agent integrations with OAuth registration, GitHub identities, Slack canvases, and custom authentication options.
* Added complete agent version history and made onboarding restartable from profile settings.
* Improved long-chat scrolling, message grouping, streaming reliability, and browser-session continuity.

## May 18th

* Added the feed, agent training and feedback, agent archiving, and restoration of earlier agent versions.
* Added browser controls and a browser side panel with presence, tabs, resize support, and conversation-specific sessions.
* Added files and artifacts panels, pasted-file support, preserved composer drafts, and full-text conversation search.
* Added workspace switching and creation, scheduled-run timezone support, agent spawning, and short-term memory across conversations.
* Added password reset, conversation recaps, improved Markdown, and clearer titles for scheduled conversations.
* Improved agent startup speed and resilience for long-running or interrupted work.

## May 11th

* Added Google and Slack sign-in, team invitations, agent integrations, managed secrets, and per-agent model selection.
* Added Slack and iMessage channels with improved threading and token handling.
* Added file uploads, attachment support, agent todo lists, remote-browser downloads, and preview environments.
* Added automatic conversation titles and a clearer chat interface.
* Improved agent startup speed from roughly 24 seconds to roughly 8 seconds in the measured path.
* Improved reliability for long chat histories, uploads, proxy requests, and recurring tasks.

## May 4th

* Added profiles, agent discovery, recent chats, and a more complete landing and authentication experience.
* Added scheduled work, persistent agent memory, and durable agent environments.
* Added an inline browser viewer and hot-reloading for agent tools and integrations.
* Rebuilt the chat interface for clearer streaming conversations.
* Improved response startup time and added the first production monitoring for agent activity.

## April 29th

* Established the initial Skydive platform.
* Added account creation, workspaces, agent creation, and direct web chat.
* Added isolated agent runtimes with persistent data and the first production web, API, and worker services.
* Added the first deployment, database, authentication, and health-check foundations required to operate Skydive.


# CLI
Source: https://getskydive.com/docs/channels/cli

Drive your agents from the terminal. They keep working when you close the laptop, and text you when they need an answer.

The Skydive CLI is how you drive your agents from the keyboard. Start work on your local repo, close the lid, and the agent keeps going in the cloud. If it hits a question, instead of stalling, you have taught your agent to text you for decisions like this. You reply, it continues. Pick up the result from your phone, Slack, or anywhere else.

```sh Binary (recommended) theme={null}
curl -fsSL https://skydive.sh | sh
skydive chat
```

```sh npm theme={null}
npm install -g skydive-cli
skydive chat
```

## Stop babysitting your terminal

With most tools, if the agent hits a question and you are not watching, the run stalls. You come back to a blinking cursor and nothing done. Claude Code is the same: it is tethered to your terminal, your laptop, and your attention. Miss a prompt, lose the run. Close your laptop lid, lose the run.

Skydive agents run in the cloud and reach you wherever you are. If the agent needs a decision, instead of stalling, you have taught it to text you. You reply, it continues. Close the laptop, the work does not stop.

**Laptop open -- start the task:**

```sh theme={null}
skydive chat --agent dev --share-machine
```

Grant local access with `ctrl+t`. The agent reads your repo through Portal, digs into the codebase, and starts on the job. You watch the diff stream in the file pane and steer it on one edge case. Then you close the lid and go to standup.

**Laptop closed -- agent still running:**

It does not need your machine anymore. It has the context it pulled and keeps working in its cloud sandbox.

It hits an ambiguous interface -- two ways to handle the retry logic, not sure which you want. Instead of stalling, you have taught your agent to text you for decisions like this: "payment retry: service layer or controller?" You reply from your phone: "service layer, keep the controller thin." It continues.

**The rest of the morning:**

**11am, Slack.** Your agent posts in `#backend` that the PR is drafted. You reply in the thread: "also add rate limiting to the new endpoint." It picks it up without you leaving Slack.

**11:30am, away from your desk.** You text it: "where are we?" It replies over SMS: payment module done, rate limiting in, tests passing, PR is open. You text back: "hand it to the PR review agent for a security pass."

It messages your code review agent -- a separate agent with its own persona and its own model, tuned for security and correctness -- and hands off the PR. The reviewer picks it up, runs through the diff, and posts its findings in Slack.

**12pm.** You read the review over lunch, make one call, and merge.

One task. A handful of messages across terminal, Slack, and SMS. The agent asked one question the whole morning and you answered it from your phone between meetings.

## What the terminal is good at

The CLI is one of several ways to reach the same agent. It has the same memory and the same conversations as web chat, Slack, and SMS, so you can start a task here and finish it anywhere. Three things it does that the other channels cannot:

> **Your local machine**
>
> `ctrl+t` grants the agent access to this machine over Portal: your repo, your `localhost`, your tools. Off until you grant it, per agent, revocable the same way.

> **Reviewing diffs in place**
>
> `ctrl+g` opens the file pane: every edit the agent made, as reviewable hunks. Comment on a line and send the comments straight back into the conversation.

> **The agent's own sandbox**
>
> `/sandbox` opens a live terminal inside the agent's cloud machine, or runs one command there and streams the output.

> **Scripts and pipes**
>
> `chat -p` sends one prompt and prints the reply. No TTY needed, so it fits CI, cron, and anything you can pipe into.

## Sharing this machine

[Portal](/docs/integrations/portal) is how an agent reaches your own computer: your repo, your `localhost`, the tools you have installed. There are two ways to open it, and they grant the same thing.

**Skydive Desktop** keeps your machine connected in the background. Approve an agent once and it can reach you whether or not a terminal is open.

**This CLI**, if you would rather not run a second app. `ctrl+t` opens Portal for the session and `/portal` grants the agent you are talking to. `skydive chat --share-machine` does both at launch, and `skydive portal open --agent <name>` runs it headless with no TUI at all.

Either way it is default-deny and per agent. Nothing reaches your machine until you say yes, and you can take it back at any time.

## Get started

> **[Install and sign in](/docs/cli/overview)**
>
> Install script or npm, the beta channel, and your first `skydive chat`.

> **[Chat TUI](/docs/cli/chat)**
>
> Keybinds, slash commands, images, the file review pane, local shell.

> **[Scripting](/docs/cli/scripting)**
>
> One-shot prompts, JSON output, and pipelines with `chat -p`.

> **[Command reference](/docs/cli/reference)**
>
> Every command, flag, and environment variable.

## Related

> **[Reaching your computer](/docs/integrations/portal)**
>
> How local machine access works and how the agent uses it.

> **[Agents overview](/docs/agents/overview)**
>
> What the cloud sandbox, memory, and connections are.

> **[Routines](/docs/agents/routines)**
>
> Agents that run without you in the loop.

> **[Platform CLI](/docs/api-reference/platform-cli)**
>
> The other CLI: the one the agent runs inside its own sandbox.


# Email
Source: https://getskydive.com/docs/channels/email

Give your agents their own inboxes

Your agents can have real email addresses. Send them requests, forward threads to handle, and get replies back over regular email. This page covers how to connect an inbox, set up your own domain, and what to do when something goes wrong.

> [!NOTE]
>
> Agents read their inboxes freely and answer questions about what's in them. They only send or reply when you ask. They never email anyone unprompted.

## Getting an inbox

Connect an inbox from the agent's profile page. You need edit rights on the agent to do this.

Pick an address from two options:

* **Skydive shared domain** (e.g. `youragent@skydivemail.ai`). Available immediately, no setup required.
* **Your own domain** (e.g. `agent@yourcompany.com`). Requires a paid plan and DNS setup. You can add that domain from this same connect dialog: open the domain dropdown and choose **Add custom domain**. See [Custom domains](#custom-domains) below.

A few things worth knowing upfront:

* Some local parts are reserved on shared domains and can't be used: `admin`, `support`, `info`, `noreply`, `security`, `billing`, `contact`, `sales`. Custom domains don't have this restriction.
* If a suggested address is taken, the platform offers the next available variant automatically. Addresses you type manually are never silently changed.
* **Addresses are permanent.** Once any workspace claims an address, it's never reissued, even after that workspace removes it. If an address shows as taken but looks available elsewhere, this is why.

Once connected, reply to the welcome email the agent sends you. It warms up the inbox so early mail from that address lands reliably rather than going to spam.

## How sending works

Receiving is always on. Sending is deliberate:

* Agents send or reply only when you ask. They won't cold-email, run outreach, or add people to lists on their own. "Grow my audience" is not standing permission to email strangers.
* When you ask an agent to email you, it already knows your address from the conversation.
* **Shared domain rate limit:** 10 outbound emails per rolling week, shared across all agents in your workspace. Replies to mail the agent received don't count against this. Inbound is unlimited. Anyone can forward to your agent or email it directly with no cap. Custom domains aren't subject to the outbound limit.
* Deliverability isn't guaranteed for brand-new inboxes. Confirm a message got through (a reply back is the clearest sign) before depending on it for anything important.
* If an outbound email bounces, the bounce details show up in the conversation so you know what happened.

### Email signature

Every email an agent sends ends with a signature block: the agent's name, its role at your workspace (for example "Executive Assistant @ Acme"), and a "Made with Skydive" line. Skydive infers the role from the agent's purpose, so a mail assistant signs as an assistant and a coding agent signs as an engineer. The signature is added after the message body, so it never touches what the agent wrote.

On paid plans you can control the signature from the agent's profile: open the agent, go to the email connection, and choose custom signature text or turn the signature off. Workspaces on the free plan get the standard signature on every send.

## What agents can do over email

> **Forward to handle**
>
> Forward a contract, a thread, or a request. The agent takes it from there.

> **Long-form requests**
>
> Detailed, async asks that don't need a live back-and-forth.

> **Attachments**
>
> Send documents, images, and files as normal email attachments. Agents read and work with them in both directions.

> **A paper trail**
>
> Every exchange lives in your inbox, threaded like any other email.

Forwarding a request works like looping in a colleague:

```text theme={null}
Forwarded: customer asking for a refund on invoice #4821.

Can you check our refund policy and draft a reply I can send?
```

## It's the same agent

Email is another doorway to the same worker. Agents bring their full [memory](/docs/agents/memory) and [persona](/docs/agents/persona) to every email thread. Skydive recognizes you by your email address, so a request sent by email connects to the same account and history as your web and Slack conversations.

## Emailing people outside your workspace

By default, only workspace members can reply to your agents in a way that reaches them. For agents that need a real back-and-forth with outside contacts (vendor coordination, client follow-up, supplier check-ins), open the agent's access settings and let people outside **Use** it.

Open the agent's **...** menu, choose **Access**, and set the outside-workspace level to **Use**. After that, anyone your agent emails can reply and the conversation continues normally.

Group threads work too. On threads with three or more participants, agents behave like a participant: they reply when addressed and don't interrupt. Replies to an agent's own private outreach are rejected rather than leaking into a group thread.

> **[Permissions & visibility](/docs/agents/permissions)**
>
> Who can see, message, and change each agent.

## Custom domains

Custom domains let agents send and receive on addresses at a domain you own. They require a paid plan. Any workspace member can add a domain; only the member who added it or a workspace admin can remove it.

You can start from two places. Both open the same setup flow:

* The agent's profile. Open the connect-inbox dialog, click the domain dropdown, and choose **Add custom domain**.
* **Workspace settings > Agent Email Domains > Add domain**. Use this when you want to manage DNS records later.

If you add a domain from the profile and automatic DNS setup doesn't finish, Skydive sends you to workspace settings so you can see the records and click Verify.

### Setup

**1. Add the domain**

From the agent's profile, open the connect-inbox dialog, click the domain dropdown, and choose **Add custom domain**. Or go to **Workspace settings > Agent Email Domains > Add domain**.

**2. Publish the DNS records**

The dialog shows the records to add at your DNS host:

* **DKIM** TXT record
* **SPF** TXT record on the `mail.` subdomain
* **MX** record (priority 10) pointing to our inbound mail host
* Bounce-handling records

If your domain already has an SPF record on `mail.`, merge ours into it rather than adding a second one. Two `v=spf1` TXT records on the same host is invalid and verification will hang.

You can also use **Entri** to publish the records automatically if your DNS host is supported.

**3. Verify**

Click **Verify domain** in the dialog. Verification doesn't start automatically. A background sweep re-checks pending domains every hour, so a DNS fix you make will heal on its own, but clicking Verify speeds it up.

**4. Create inboxes**

Once the domain shows as **Verified**, you can create agent inboxes on it.

### Verification statuses

Verification moves through these states: `not started > pending > verifying > verified` (or `invalid` / `failed` if a record is wrong or missing). Each DNS record shows its own badge so you can see exactly which one is holding things up.

DNS propagation can take a few minutes to a few hours. If verification still hasn't succeeded after 48 hours, you'll get a failure notice by email. A successful verification also sends a confirmation. Re-running Verify resets the 48-hour clock.

### After verification: delivery gaps to check

"Verified" confirms the records the platform checks. Additional issues can still affect delivery, and the dashboard warns about the first two after verification:

* **Missing apex SPF.** Our SPF record lives on `mail.` (the sending subdomain), but your From address is on the apex (`agent@yourcompany.com`). Strict receivers that only check the apex can rate-limit or reject. Fix: add `include:amazonses.com` to your apex SPF, or a dedicated apex SPF entry.
* **Duplicate DMARC.** Two DMARC TXT records at the apex is undefined behavior; some receivers reject all mail from the domain. Keep exactly one.
* **A competing MX at higher priority (Google Workspace).** If your domain also runs Google Workspace, its MX (usually `smtp.google.com`, priority 1) outranks Skydive's inbound MX, so Gmail receives all inbound mail first. If the address isn't a recognized Google account, Gmail rejects it with `550 5.1.1` and Skydive's inbound host never sees the mail. Verification still passes, because our records exist. Fix: complete the [Google Workspace coexistence](#google-workspace-coexistence) setup below.

### Google Workspace coexistence

You can run agent email on a domain that also runs Google Workspace. Google handles recognized users; agent addresses are unrecognized accounts that Google forwards to Skydive's mail host.

If you'd rather skip the routing setup, put agent email on a dedicated subdomain (e.g. `mail.yourcompany.com`) instead.

**1. Add the domain**

Go to **Workspace settings > Agent Email Domains > Add domain**. If your domain's MX already points at Google, the platform registers the domain for coexistence automatically and no conflict error appears.

> [!WARNING]
>
> Verification will pass even though inbound mail bounces until you finish the routing setup below. The domain checks confirm Skydive's records exist, not that they take priority over Google's. Don't stop at "Verified".

**2. Publish DNS records**

Add the records the dialog provides, keeping your existing Google Workspace MX records in place. Google's MX should stay at priority 1; add Skydive's inbound MX at priority 10. Also add the DKIM record and the `mail.` subdomain bounce record.

Example for a domain `yourcompany.com`:

| DNS name | Priority | Mail server |
| - | - | - |
| `yourcompany.com` | `1` | `smtp.google.com` |
| `yourcompany.com` | `10` | The inbound MX host from the add-domain dialog |
| `mail.yourcompany.com` | `10` | Bounce-handling host |

MX priority alone doesn't route unrecognized addresses to Skydive. You also need the routing rule below.

**3. Add Skydive as a mail host in Google Workspace**

In the Google Workspace admin console, go to **Apps > Google Workspace > Gmail > Hosts** and click **Add Route**:

* Name: `Skydive`
* Select **Single host**
* Host name: the inbound MX host from the add-domain dialog (the same value as your priority-10 MX record)
* Port: `25`
* Enable the recommended TLS options

**4. Add a routing rule**

In Gmail settings, scroll to **Routing** and click **Add another rule**.

* Check **Inbound** and **Internal - Receiving**
* Set the action to **Modify message**
* Check **All inactive and unrecognized accounts**
* Check **Change route** and select the Skydive route

Save the rule. Gmail will now forward mail for any address that isn't a recognized Google Workspace account to Skydive.

**5. Create inboxes before mail flows**

Create each agent inbox address before mail starts flowing to it. Google forwards but doesn't queue for unknown addresses, so mail to an address with no inbox yet will bounce.

> [!NOTE]
>
> The routing rule only forwards addresses Google doesn't recognize. If you later create a Google account or alias for an agent address, Gmail stops forwarding it to Skydive and that address goes dark. Keep agent addresses out of Google Workspace.

> **[Reach it on the go too](/docs/channels/text)**
>
> Add text messaging for quick, mobile-friendly check-ins.


# Overview
Source: https://getskydive.com/docs/channels/overview

Reach your agent everywhere you already work

A channel is where you talk to your agent: web chat, Slack, email, or SMS/iMessage. The same agent is reachable on all of them, with **one identity and one memory** everywhere.

## Available channels

> **[Web chat](/docs/channels/web)**
>
> The full-featured workspace in your browser. Files, history, and live previews.

> **[Slack](/docs/channels/slack)**
>
> DM your agent or @-mention it in channels alongside your team.

> **[Email](/docs/channels/email)**
>
> Give your agent its own inbox. Forward it things, get replies back.

> **[SMS & iMessage](/docs/channels/text)**
>
> Text your agent from your phone.

On a Mac, the [desktop app](/docs/desktop/mac) is web chat plus your local files and apps. In a terminal, the [Skydive CLI](/docs/cli/overview) is the same agents at your prompt.

## One agent, every channel

Channels are doorways to the same worker:

* **Shared memory**: teach it something in Slack and it remembers in email and web chat.
* **Unified identity**: Skydive recognizes you by your Slack handle, email address, or phone number, and ties the conversation to your account.
* **The right channel for the moment**: start a deep task in web chat where you can watch it work, then check in from Slack or text.

## Choosing a channel

| Channel | Best for |
| - | - |
| **Web chat** | Heavy lifting: file work, live app previews, long sessions, watching the agent work step by step. |
| **Slack** | Team collaboration: shared context, quick asks, @-mentioning the agent into a thread. |
| **Email** | Async handoffs: forwarding documents, long-form requests, a paper trail. |
| **SMS / iMessage** | On the go: quick questions, reminders, nudges from your phone. |

## Attachments work everywhere

Send files on any channel: documents, images, spreadsheets, audio. The agent fetches, reads, and works with them, then returns results you can download. See [Attachments](/docs/capabilities/files#attachments).

> **[Add your agent to Slack](/docs/channels/slack)**
>
> The fastest way to put an agent in front of your team.


# Slack
Source: https://getskydive.com/docs/channels/slack

Install Skydive in your Slack workspace and make agents with their own Slack identity.

Adding an agent to Slack takes one click from the [Skydive app](https://skydive.com). Your agent joins the Slack workspace as its own Slack app, with its own name and avatar, ready for DMs and channel @-mentions.

Why teams add their agents to Slack:

* **The whole team can use it.** Anyone in the Slack workspace can DM the agent or @-mention it in a channel. No Skydive accounts needed.
* **It works where the conversation already is.** The agent reads the thread it is pulled into and answers with that context.
* **It appears as its own app.** Each agent shows up under its own name and avatar.

Need help at any point? Email [support@skydive.com](mailto:support@skydive.com) and we'll help.

![Emma messaging from Slack](https://mintcdn.com/skydive/2VR7bKqFZZqW4hyR/images/agents/slack-emma.png?fit=max\&auto=format\&n=2VR7bKqFZZqW4hyR\&q=85\&s=3483c3949741dbe5a0047c5d5dd2a368)

## What gets installed

The first time you add an agent to a Slack workspace, two things install:

* **The Skydive app**, once per Slack workspace. It handles sign-in, general Skydive questions, and creating agent apps. This is the app listed in the Slack Marketplace.
* **An app for your agent.** Each agent you add gets its own Slack app, named after the agent and carrying its avatar. Your team DMs and @-mentions the agent app, not the Skydive app.

During the install, Skydive asks for permission to create and install agent apps. That permission is what lets you add more agents later without going through Slack's install flow each time.

## Add an agent to Slack

You can start from three places in Skydive. All of them run the same install.

### From a new chat

**1. Click 'Add me to Slack'**

Open a fresh conversation with the agent. The channel options sit above the message box.

**2. Approve the install**

A Slack window opens. Pick the Slack workspace and approve the permissions.

**3. Message on Slack**

The button becomes **Message on Slack** and takes you straight to the agent's DM.

### From the agent's profile

**1. Open the Connections card**

On the agent's profile, find the **Slack** row.

**2. Click 'Add' and approve**

Approve the install in the Slack window that opens.

**3. Manage from the same row**

Once connected, the row shows the Slack workspace name and a **Manage** button.

### From a conversation

Tell the agent "add yourself to Slack" and it brings up the install in the chat.

**1. Ask, or look for the card**

Ask the agent to add itself to Slack, or wait for it to suggest it. Either way, a card appears in the conversation: **Add \[agent name] to Slack**.

**2. Click 'Add' and approve**

Same install, no need to leave the chat.

> [!NOTE]
>
> If your Slack workspace requires admin approval for new apps, Slack shows a "request to install" screen instead. Send the request. Once a Slack admin approves Skydive, run the install again from any entry point above.

## Talk to your agent in Slack

Once added, the agent works in Slack like any member:

* **DM it** for private work: "pull yesterday's signups and send me the list."
* **Invite it to a channel**, then @-mention it in threads: "@agent summarize this thread and post the action items, with owners." It replies in-thread with the conversation's context in view.
* **Share files with it** in Slack. It picks them up, works with them, and replies with results.
* **Send tables.** If you use Slack's table feature, the agent reads the table's actual contents, not a placeholder.
* **Same agent everywhere.** It has the same [memory](/docs/agents/memory) in Slack, web chat, and email. What it learns in one place carries to the rest.
* **Tell it to mute itself.** In a channel or thread, ask the agent to stay quiet when you do not need it chiming in on every message.

Put it on a [routine](/docs/agents/routines) to post standups, digests, and reminders to a channel on a cadence.

## How the agent writes on Slack

Slack is a casual, fast channel, and the agent writes accordingly. By default it keeps replies short: one or two sentences, plain text, no preamble or recap. It leads with the answer and stops.

When you ask for something that genuinely warrants a longer reply, the agent goes long. It uses Slack formatting (code snippets, an occasional bold phrase) only when it helps, not to decorate a conversational answer.

This keeps DMs and channels readable, the same way a good teammate on Slack would.

## Reply threading in DMs

By default, the agent replies to each message in a Slack DM as a new top-level message. You can switch to **threaded replies** so the agent responds inside the same thread as the message it's answering, keeping your DM history easier to skim.

There are two independent toggles, each covering a different DM surface:

| Setting | What it controls |
| - | - |
| **1:1 DMs** | Standard direct messages between you and the agent |
| **Group DMs** | Multi-person direct message conversations the agent is part of |

To change either one, open the agent's profile, find the **Slack** row, click the **···** menu, and choose **Slack threading**. Each toggle saves immediately.

## Mute an agent in a thread or channel

Agents can get noisy in busy channels. In a **channel or group thread** (not a 1:1 DM), you can tell the agent to mute itself for that conversation:

```text theme={null}
Stop jumping in on this thread unless I @ you.
```

```text theme={null}
Mute yourself here for a bit.
```

What happens:

1. The agent goes **silent in that conversation only**. It does not keep running on every new message in the thread or channel while muted.
2. **@-mention the agent again** to bring it back. The mute lifts as soon as someone tags it, and it responds as usual.
3. Other conversations are unchanged. Muting in one thread does not mute DMs or other channels.

Use this when a thread has become a human discussion, a standup the agent already handled, or anything where you want the agent on-call but not speaking until asked.

> [!TIP]
>
> If an agent "stopped responding" in a channel where it used to reply, check whether someone asked it to stay quiet. @-mention it once to unmute and test.

## The Skydive app

The Skydive app is more than the installer. It is an assistant for your whole Slack workspace. DM it directly.

Use it to:

* **Create new agents.** Tell it the job ("we need an agent that triages #support") and it sets one up.
* **Get help with existing agents.** Ask what agents your team has, what they do, or why one isn't responding.
* **Learn Skydive.** How agents work, what they can do, what to try next.
* **Manage your Skydive workspace.** Questions about members, settings, connections, or billing: it points you to the right place.

Anyone in the Slack workspace can message it, whether or not they have a Skydive account yet.

## Manage your Slack workspaces

Slack connections for your whole Skydive workspace live in [**Profile & Settings**](https://skydive.com/profile) in the Skydive app, under **Slack**:

* **Install** / **Add workspace**: run the Slack install for a new Slack workspace. An agent can be in more than one Slack workspace, and a Slack workspace can hold many agents.
* **Disconnect**: remove Skydive's access to that Slack workspace. Agent apps already in it stop working until you install again.

Each agent's own connections live on its profile, behind **Manage** on the Slack row:

* **Remove**: take the agent out of one Slack workspace.
* **Connect to another workspace**: add the agent to an additional Slack workspace.

Your workspace settings also have a **Slack apps** section listing every agent Slack app installed across your Skydive workspace. Click **Uninstall** on a row to remove that agent's app from its Slack workspace. The agent stops responding there. You can reconnect the agent to Slack later from its profile.

## Troubleshooting

> **The agent isn't replying in a channel**
>
> Check these in order:
>
> 1. The agent's app is invited to the channel (invite it like a person).
> 2. The sender has [Use access or above](#who-the-agent-answers). Without it, the agent stays silent with no error.
> 3. The agent may be **muted in that conversation**. If someone told it to stay quiet, @-mention it once to unmute and try again.

> **How do I get a muted agent talking again?**
>
> @-mention the agent in that same channel or thread. The mute is for that conversation only, and an @-mention lifts it immediately. You do not need to reinstall or reconnect Slack.

> **The install didn't finish or permissions were declined**
>
> Run the install again from [**Profile & Settings**](https://skydive.com/profile) → **Slack** and approve everything Slack asks for. Skydive needs the app-management permission to create your agent's app; declining it fails the install with "Slack didn't grant the permissions needed."

> **Someone removed the Skydive app from Slack**
>
> Uninstalling from the Slack side quietly removes the connection in Skydive too. Reinstall from [**Profile & Settings**](https://skydive.com/profile) → **Slack**, then re-add each agent from its profile.

> **An agent's app was removed from Slack**
>
> Re-add the agent from any entry point above.

> **The install window closed or was blocked**
>
> Allow pop-ups for the Skydive app and click the install button again.

## For Slack workspace admins

If your Slack workspace requires admin approval for new apps, here's what to expect:

**1. A member requests Skydive**

When someone on your team clicks Add to Slack, Slack sends you their request through your usual app-approval flow.

**2. Approve the Skydive app**

Review and approve **Skydive** in Slack's app management settings, the same place you approve any app.

**3. Set an automation rule for agent apps**

Skydive creates one Slack app per agent your team adds, each named after its agent. When you approve Skydive, Slack offers to create an automation rule that auto-approves these agent apps, so each new agent doesn't file its own approval request. You control the rule: scope it to the conditions you want, add your own conditions, or remove it entirely. Slack checks the rule on every install.

A few facts that matter for review:

* Each agent app is created and managed by the approved Skydive app. It requests only the permissions agents need and can't be installed outside your Slack workspace.
* Agent apps appear on Slack's app management page like any other app. Removing an agent in Skydive removes its app; uninstalling Skydive from Slack shuts off every agent app.
* Slack marks these apps as managed in its app settings pages. Leave their configuration to Skydive: editing an agent app's settings directly in Slack can break the agent.

## Privacy and permissions

What an agent can do in Slack, at a glance:

| Capability | What it does | Makes changes? |
| - | - | - |
| Send messages | DM you, reply in threads, post to channels it's in | Yes |
| Read messages | See conversations in the DMs and channels it's part of | No |
| Join channels | Enter channels it's invited to | Yes |
| Files and reactions | Pick up shared files, add emoji reactions | Yes |
| Create agent apps | Skydive app only: provision each agent's own app | Yes |

Agents only see conversations they're part of.

### Who the agent answers

Each agent carries [access settings](/docs/agents/permissions) in Skydive. They apply in Slack too. Skydive recognizes each sender by their Slack identity and checks their access before the agent replies:

| Access | Who gets a reply in Slack |
| - | - |
| **No access** | Nobody (except the agent's owner and workspace admins) |
| **View** | Nobody: seeing and messaging are separate |
| **Use** (default) | Workspace members and anyone whose Slack identity matches a grant |
| **Edit** or **Owner** | Everyone above |

When someone without access messages the agent, it stays silent: no reply, no error. The same rule applies per person in channels. An agent @-mentioned in a thread answers the people who have access and ignores the rest. If a teammate reports the agent "not responding," check the agent's access settings first.

## Questions

**Do my teammates need Skydive accounts?**
No. Anyone in the Slack workspace can talk to an agent whose access settings allow it. Skydive recognizes each person by their Slack identity.

**Can I add one agent to several Slack workspaces?**
Yes. From the agent's profile, click **Manage** on the Slack row, then **Connect to another workspace**.

**Something else?**
Email [support@skydive.com](mailto:support@skydive.com). Include what you tried, what happened, and the agent's name.


# iMessage
Source: https://getskydive.com/docs/channels/text

Text your agent from your phone

SMS and iMessage make your agent reachable from your phone. Use it for quick asks, reminders, and nudges when you're away from a screen.

## How it works

Your agent has a phone number. Send it a text and it responds in the same thread, with its full [memory](/docs/agents/memory) and [persona](/docs/agents/persona). It is the same agent you talk to in web chat and Slack.

> **Quick questions**
>
> "What's on my calendar tomorrow?" gets a fast answer without opening an app.

> **[Reminders & nudges](/docs/agents/routines)**
>
> "Remind me at 5 to call the supplier." The agent texts you when it's time.

> **On-the-go handoffs**
>
> Kick off a task by text and let the agent work while you're moving.

> **Photos & files**
>
> Text a photo or file and the agent reads and acts on it.

## Connecting iMessage

You need [Edit](/docs/agents/permissions) on the agent. Open its profile and click **Add** next to iMessage. Skydive will text a code to your phone number. Enter the code (it submits automatically on the last digit), then pick your line type.

The phone you verify becomes the number Skydive uses to confirm who you are across SMS conversations. If it's already verified on your account, the step is skipped. Who can later text the agent follows the same [Access](/docs/agents/permissions) settings as every other channel.

You'll get a card with the agent's number and a first message. Send that message from the phone you verified. That's the whole setup.

### Android

Same flow. Android just can't tap the card to send, so copy the number and the first message and text them yourself.

## Shared vs. dedicated lines

When you connect an agent to iMessage, you pick how it gets its number.

| | Shared | Dedicated |
| - | - | - |
| **Cost** | Free | \$100 / month per agent |
| **Number** | From a pool shared across workspaces | Exclusively this agent's own number |
| **Turns on** | When you send the first text from your verified phone | When you buy it. You can text immediately. |
| **Who can reach it** | People already tied to this agent on this line (you, after connect; anyone who sends the reconnect code) | Anyone who texts the number. The agent's privacy setting then decides whether they get in. |
| **Sending capacity** | Shared with the pool | Full capacity, just this agent |
| **Limit** | Your workspace can hold a limited number | No limit |

Most workspaces start on a shared line. If an agent is sending at high volume, or you want a number that is permanently and exclusively that agent's, a dedicated line is the upgrade.

Shared lines have a per-workspace sending cap. When an agent hits the cap mid-turn, it sends one friendly notice instead of going silent, then stops delivering for that turn. The notice includes a link to upgrade your plan. Dedicated lines have their own capacity and no shared cap.

Agents on text are responders. They reply in threads they are already in. They are not built for cold outreach, and they cannot text a number that has never written them.

If you want the agent to answer the public, give it a dedicated line and set Access to **External**. Then whoever texts that number gets a reply. Keep it private if it should only talk to your team.

## Group texts

Add the agent to a group the way you'd add anyone else. Ask it for its number, then add that number. It cannot start a group, and it cannot add or remove people.

In a group, the agent reads the thread and decides when to jump in. One-on-one, it always replies.

What happens next follows [Access](/docs/agents/permissions) for people outside your workspace:

* **External**: the agent joins. Anyone in the group can talk to it, including people without a Skydive account.
* **Private**: it only joins if at least one member is already allowed to use it. If nobody in the group has that access, it stays out and the group never hears from it.

## What agents can send

Text replies go out automatically. Agents can also send photos, video, and files, a link with a preview, a voice memo, their contact card, and tapbacks (love, like, laugh, and the rest).

## It knows it's you

Skydive recognizes you by your phone number and ties the text conversation to your account. What you discuss over text connects to the same history and memory as every other channel.

A text exchange looks like any other:

```text theme={null}
You:    Did the Acme contract get signed yet?
Agent:  Not yet. It's been with their legal team since Tuesday.
        Want me to send a nudge?
You:    Yes please
```

## Best for the small stuff

Text works best for short, immediate interactions. For heavier work (building something, processing files, watching the agent run), switch to [web chat](/docs/channels/web). The conversation carries over: one agent, one memory.

If a run fails over text, the notice is shorter and in the agent's voice. The same errors are decoded in [Troubleshooting](/docs/get-started/troubleshooting-errors#text-message-failures-imessagesms). In a group, the agent just stops.

> **[See all the ways to reach your agent](/docs/channels/overview)**
>
> Web, Slack, email, and text: one agent across all of them.


# Web
Source: https://getskydive.com/docs/channels/web

The full-featured agent workspace in your browser

Web chat is the richest way to work with your agent. Watch it work in real time, hand it files, and see everything it produces.

On a Mac, the [desktop app](/docs/desktop/mac) is this same interface plus access to your local files and apps.

## Mobile

Skydive works as a progressive web app on iPhone. Open `skydive.com` in Safari, tap **Share**, then **Add to Home Screen**. The app installs without going through the App Store, launches full-screen, and stays in sync with your other sessions.

A native iOS app is planned for the future.

## What you get

> **Live progress**
>
> Watch the agent plan and act step by step. Multi-step jobs pin a Plan card above the composer.

> **File handoff**
>
> Drop in documents, images, spreadsheets, and audio. Download whatever the agent creates.

> **[Shared files](/docs/capabilities/files)**
>
> A Files tab where you and the agent both read and write, and where deliverables show up.

> **[Live app previews](/docs/capabilities/webserver)**
>
> When the agent builds a web app, view and click through it inline.

> **[Shared browser](/docs/capabilities/browser)**
>
> See the agent's browser, and step in to log in or solve a captcha when needed.

> **History**
>
> Revisit any past conversation with the agent.

## Sending attachments

Use the composer to attach files. The agent downloads and works with them (reading a contract, transcribing audio, analyzing a spreadsheet) and returns results you can download from the chat.

## Media in replies

When an agent shares images, GIFs, or videos, they render inline as a gallery rather than as download cards. Up to ten items appear in a responsive grid; a `+N more` label collapses anything beyond that. Click any tile to open it full-size. If you want a downloadable file card instead, the agent can pass `--download` to `platform web attachments share`.

## Diagrams

Mermaid diagrams render inline in web chat. When an agent responds with a fenced code block tagged ` ```mermaid `, the chart draws directly in the message rather than showing raw text. Flowcharts, sequence diagrams, entity-relationship diagrams, and every other Mermaid diagram type are all supported. The same rendering applies to messages you send, assistant replies, and the transcript viewer.

## Conversations

Each chat is its own thread. All threads share the agent's [memory](/docs/agents/memory). Start a new conversation for a new topic; the agent still carries everything it has learned.

### Archive and delete

Archive a finished thread to clear it from your main list without losing it. Archived chats move to an **Archived** section in the sidebar and become read-only. Restore one at any time to pick up where you left off. Archiving is private to you: in a shared chat, everyone else still sees it as active.

Deleting is permanent, so Skydive asks you to confirm first. In a shared conversation, you can only delete it when you are the last person in it.

### Context compaction

When a conversation runs long, the agent automatically compacts earlier context to stay within its working window. Nothing is lost: the agent carries a summary of what was compacted forward, and the full transcript stays in your history.

Web chat surfaces compaction inline. A `✓ Compacted this conversation` marker appears in the timeline at the point where compaction happened, with a timestamp. While compaction is in progress you'll see a live indicator. No action is needed on your part.

### Fork a chat

Fork a chat to copy the whole thread into a new conversation and branch off without touching the original. The copy carries the full history up to that point, so the agent keeps all the context, and you can take it in a different direction while the source stays as it was. Find it in the conversation options menu (the dots next to the title), alongside **Rename chat**.

A fork also frees a thread from its channel. Forking a Slack conversation gives you a copy you can carry on in web chat, which is handy when you want to keep working from the browser on something that started in Slack.

## More than one agent in a chat

A web conversation can hold up to three agents. When work needs a specialist or a second opinion, an agent can pull a peer into the same thread. The peer sees the conversation and takes part directly, alongside you. Added and removed agents are announced in the thread, so it is always clear who is present. You can also ask an agent to bring in or drop another agent. This applies to web chat only; Slack, email, and text have their own membership.

## Switching channels

When you are chatting in web and want to continue on iMessage, Slack, or email, look for the channel pill at the top of the conversation. Connected channels show an **Open in \[channel]** action that takes you straight into that thread.

## When a run fails

Most of the time a failed run is self-explanatory: the agent tells you, you retry, it works. Two kinds of failure are worth understanding.

**Usage and limits.** If your workspace is out of credits or over a rate limit, the error says so and links to [usage and billing](/docs/account/usage). Top up or wait for the limit window to reset.

**Model provider errors.** Sometimes the model behind the agent is what fails, and the error names the fix when there is one. The one to know today: Claude subscription connections refuse requests until model training is verified off. The fix is in Claude, not Skydive: open Claude's **Settings > Privacy** and turn off **Help improve our AI models**, then retry. The error copy shows the same steps.

Where the message appears depends on where you're talking to the agent: live in the conversation as it fails, in [run history](/docs/workspaces/feed) afterwards, and in a failure notice on iMessage threads. The copy is the same everywhere, so a fix you look up once applies anywhere the error catches you.

Providers add new failure modes over time. When the error copy already names a fix, follow it. When it's the generic "model stopped responding" line, retry once; if it keeps failing, the problem is the provider connection, not your prompt.

## When to use web chat

Use web chat when the work is substantial: building something, processing files, a long back-and-forth, or anything where watching the agent work helps. For quick asks, [Slack](/docs/channels/slack) or [text](/docs/channels/text) may be faster.

> **[Bring it to your team](/docs/channels/slack)**
>
> Add the same agent to Slack so everyone can reach it.


# Authentication
Source: https://getskydive.com/docs/cli/authentication

Two sign-ins for two jobs: an API key for management, a browser session for chat

The CLI has two credentials, each serving a different job:

| | API key | Chat session |
| - | - | - |
| **Command** | `skydive auth login` (the default) | `skydive auth login --web` |
| **Used by** | `agents`, `keys`, `secrets`, `workspace`, `chat` | `chat`, `conversations`, `messages` |
| **Identity** | An agent-scoped key (`sky_live_...`) | You, your whole account |
| **How you get it** | [skydive.com/settings/account](https://skydive.com/settings/account) or `skydive keys` | Browser device sign-in |

An API key now works for `skydive chat` too. If you have only a key (no browser session), the TUI opens, you can send messages, and the conversation is attributed to the key's owner. A few features that require a full account session are soft-disabled with a prompt: workspace switching (`ctrl+w`), Portal machine sharing (`ctrl+t`), and `--share-machine`. For those, run `skydive auth login --web` once to add a session alongside the key.

Most people who want the full surface set up both once. Both are stored in `~/.config/skydive/config.json`.

## Sign in for chat (`--web`)

```bash theme={null}
skydive auth login --web
```

A browser page opens for a one-time device sign-in. Approve it and the CLI stores a session on this machine. If you skip this, `skydive chat` prompts you the first time it needs it.

`--web` requires an interactive terminal. For CI, a pipe, or a remote box, set the session token directly:

```bash theme={null}
export SKYDIVE_SESSION_TOKEN=<token>
```

## Sign in with an API key (the default)

```bash theme={null}
skydive auth login
```

With no flags, `auth login` starts the API-key flow. It prompts:

```text theme={null}
Enter your API key (from skydive.com/settings/account):
```

Paste the key and it is validated against the API and saved. To skip the prompt, pass the key inline or set it in the environment:

```bash theme={null}
skydive auth login --api-key sky_live_xxxxxxxx
export SKYDIVE_API_KEY=sky_live_xxxxxxxx
```

Keys start with `sky_live_`. Create and revoke them at [skydive.com/settings/account](https://skydive.com/settings/account) or with [`skydive keys`](/docs/cli/reference#keys).

> [!NOTE]
>
> An empty paste, or a key without the `sky_live_` prefix, fails with an error. If you meant to sign in for chat, use `--web`.

## Check and clear state

```bash theme={null}
skydive auth status   # shows both credentials' state
skydive auth logout   # clears both: API key and chat session
```

> [!WARNING]
>
> Treat API keys and session tokens like passwords. Don't commit them, and revoke a key immediately if it leaks.

## Workspaces

Commands act on your **active workspace**. If you belong to several, list and switch:

```bash theme={null}
skydive workspace list
skydive workspace switch acme-team
```

The switch persists across commands until you switch again. Run `skydive workspace switch` with no argument to pick interactively.


# Chat TUI
Source: https://getskydive.com/docs/cli/chat

The interactive terminal chat experience

`skydive chat` opens a full-screen terminal chat with your agents. It streams replies live, renders Markdown, shows readable diffs for file edits, and keeps your conversation history one keystroke away.

```bash theme={null}
skydive chat                 # pick an agent, then a conversation
skydive chat --agent grace   # skip the picker, open one agent's conversations
```

## Moving around

The TUI has three screens: a welcome screen (pick or create an agent), a conversation home, then chat. `esc` goes back one level.

Press `?` in chat (with an empty composer) for the full searchable keybind reference. The essentials:

| Keys | Action |
| - | - |
| `↵` | Send message. Steers the agent mid-run. |
| `shift+↵` | Insert a newline |
| `↑` / `↓` | Recall prompt history (caret at line start or end) |
| `esc` | Cancel the latest steer (running) or go back (idle) |
| `ctrl+c` | Cancel the run. Twice quits. |
| `ctrl+g` | Toggle the [file review pane](#the-file-review-pane) |
| `ctrl+s` | Toggle the [side panel](#the-side-panel) |
| `ctrl+p` | Switch model for this conversation or as the new agent default |
| `ctrl+o` | Switch theme |
| `ctrl+t` | Open or close Portal on this machine |
| `ctrl+l` | Open this conversation in the browser |
| `ctrl+r` | Act on a waiting connect or ask card |
| `ctrl+a` | Archive this conversation (composer empty) |
| `ctrl+f` | Fork a channel conversation into a chat you can continue here |
| `pgup` / `pgdn` | Scroll the transcript |
| `end` | Jump to the latest message when scrolled up |

The bottom-right corner of the screen shows the active model. When a conversation is pinned to a model that differs from the agent's default, the footer names both: the pinned model for this thread and the agent's usual default. Switching via `ctrl+p` lets you scope the change to just this conversation or move the agent's default for all future conversations.

Pinned conversations stay in reach. In the conversation picker (the screen you pick a thread from when you run `skydive chat` or `/conversation`), press `ctrl+p` on a row to pin or unpin it. Pinned conversations sort to the top of the list and get a pin marker. Pinning is per person, not per agent.

## Send while it works

You don't have to wait for a run to finish. Send another message mid-run to **steer** the agent: it folds your input into the work in progress. `esc` cancels the latest steer.

## Images

Attach images three ways:

* **Paste** from the clipboard with `ctrl+v`
* **Drag and drop** image files onto the terminal
* `ctrl+x` drops the last staged image before sending

## Slash commands

Type `/` in the composer to open the command menu. Type to filter, `↵` to run.

| Command | What it does |
| - | - |
| `/sandbox [command]` (or `/sbx`) | Run a command in the agent's sandbox. No command opens a terminal. |
| `/new` | Start a new conversation with this agent |
| `/fork` | Fork this conversation into a new one (full copy) |
| `/conversation [id or link]` (or `/conv`) | Switch conversation. Without an id, pick from a list. |
| `/search [query]` (or `/find`) | Search conversations by title, message, or participant |
| `/rename <new title>` (or `/title`) | Rename this conversation, for everyone |
| `/archive` | Archive this conversation and hide it from your list |
| `/model` | Switch the model for the next run |
| `/effort` (or `/reasoning`) | Change the current model's reasoning effort |
| `/theme` | Switch the color theme |
| `/status` | Show connection and session status |
| `/computer [restart]` | Show the agent sandbox's CPU, memory, and disk. `restart` reboots it. |
| `/prs` (or `/pr`) | List this conversation's linked PRs, then open or copy one |
| `/plan [show\|hide\|minimize]` (or `/todo`) | Show, hide, or minimize the pinned plan card to a header row. Click the card header to toggle. `/plan collapse` and `/plan expand` fold and unfold completed task rows inside the card. When you reopen a conversation whose plan is already finished, the card loads minimized automatically. |
| `/subagent [minimize]` | Minimize or restore the pinned subagent card. Click the card header to toggle. Finished subagent cards also load minimized when you reopen a conversation. |
| `/share` | Open or close Portal on this machine |
| `/portal` (or `/local`) | Give this agent access to your machine over Portal |
| `/browser` | Open this conversation in the browser |
| `/connect` | Act on a waiting connect card |
| `/export [md\|txt]` | Export this conversation to a markdown or text file in your current directory |
| `/copy` (or `/cp`) | Copy the agent's last message to the clipboard |
| `/compact` | Summarize older turns now to free up context |
| `/help` | Show keybinds and commands |
| `/exit` (or `/quit`) | Quit the chat |

## The side panel

Press `ctrl+s` to open a right-hand panel beside the chat. It shows at a glance what your agent is doing and the resources it is using, without leaving the conversation.

The panel is hidden by default and only appears when your terminal is wide enough to fit it alongside a readable chat column (72 columns minimum). Toggling on a narrower terminal shows a brief warning instead of squeezing the layout.

**What the panel shows:**

* **Conversation and agent** at the top: the conversation title and the active agent name.
* **Computer stats:** live CPU, memory, and disk usage for the agent's sandbox, updated every 30 seconds. Shows an idle line when no sandbox is running and a stale-data note if the last sample is more than five minutes old.

## The file review pane

Open it with `ctrl+g`. The file pane sits alongside the chat, below it, or fullscreen. The layout adapts to terminal size and is mouse-resizable. Two tabs:

**Changes** shows every edit the agent made in this conversation as reviewable diff hunks. Toggle between the full session and the last turn with `t`. Leave comments on any line or selection and send them back to the agent with `s`.

**Files** browses the agent's `~/Documents/` folder. Navigate the tree and preview text files.

| Key | Action |
| - | - |
| `1` / `2` | Switch Changes / Files tabs |
| `j` / `k` or `↑` / `↓` | Move in the focused pane |
| `↵` / `o` / `l` | Focus the file preview |
| `h` / `←` | Return focus to the file navigator |
| `/` | Search files in the active tab |
| `ctrl+d` / `ctrl+u` | Half-page down / up |
| `]` / `[` | Next / previous edit hunk (Changes) |
| `f` / `F` | Next / previous changed file (Changes) |
| `t` | Toggle session / last-turn scope (Changes) |
| `v` | Start or clear a line selection |
| `c` | Comment on selection or current line |
| `e` / `d` | Edit / delete comment at cursor |
| `s` | Send all comments into the composer |
| `y` | Copy comments to clipboard |
| `\` | Toggle side / stacked layout |
| `esc` | Preview, then tree, then close |
| `q` or `ctrl+g` | Close the pane |

## Run local shell commands

Start a message with `!` to run it in your local shell instead of sending it to the agent:

```text theme={null}
! git status
! cd ~/dev/project && ls
```

Commands run with your own shell environment (aliases, functions, PATH), and the working directory carries over between `!` commands. Output renders in the transcript, so you can reference it in your next message to the agent.

## Give the agent local access

The agent runs in its own cloud sandbox. To let it run commands on **your** machine for a task, press `ctrl+t` in chat and approve the prompt (`y` to allow, `n` to keep the machine private). Access is default-deny, per agent, and revocable the same way. You can also start chat with the grant in place:

```bash theme={null}
skydive chat --share-machine
```

See [Reaching your computer](/docs/integrations/portal) for how local access works.

## Connect cards

When an agent needs a credential or a service authorization, a **connect card** appears in the transcript. Press `ctrl+r` to run the newest card: sign in via the browser, or enter the credential in the prompt that opens.

## Themes and notifications

* The TUI follows your terminal's light or dark background. Pin a colorscheme with `ctrl+o`, `--theme <id>`, or the `SKYDIVE_THEME` environment variable.
* When a run finishes or needs input while the terminal is unfocused, the CLI shows a desktop notification. Disable with `--no-notify`.

## The welcome screen and conversation home

**Welcome screen** is the first thing you see when you run `skydive chat` with no agent specified. Pick an existing agent from the list, or press `ctrl+n` to create one.

| Key | Action |
| - | - |
| `↑` / `↓` | Move selection |
| `pgup` / `pgdn` | Move a page at a time |
| `↵` | Open agent |
| `tab` | Toggle mine / all agents |
| `ctrl+n` | Create a new agent |

**Conversation home** opens once you pick an agent (or when you use `--agent`). It mirrors the web app home: a composer at the top and your conversation list below.

* Type a message and press `↵` to start a fresh conversation seeded with it.
* Press `↵` on an **empty** composer to open the highlighted conversation.
* Type `/query` in the composer to search conversations by title, message body, or participant name. The results filter as you type.

Unread conversations are flagged. Opening one marks it read here and in the web app.

| Key | Action |
| - | - |
| `↑` / `↓` | Move selection |
| `pgup` / `pgdn` | Move a page at a time |
| `↵` | Open conversation (empty composer) or send as new conversation (with text) |
| `ctrl+a` | Archive (hides it for you only; again to restore) |
| `ctrl+d` | Delete for everyone (`y` or `↵` to confirm) |
| `tab` | Show / hide the agent's own runs (schedules, webhooks, subagents) |
| `shift+tab` | Show / hide conversations you archived |
| `esc` | Back to the welcome screen |

To read a conversation without the TUI, use [`skydive conversations show`](/docs/cli/reference#conversations).

> [!NOTE]
>
> The TUI runs on the Bun runtime. The CLI provisions a private, pinned Bun automatically on first run; nothing to install. If your machine can't fetch it (offline or unsupported platform), install [Bun](https://bun.sh) yourself or use `chat -p`, which runs under Node.


# Skydive CLI
Source: https://getskydive.com/docs/cli/overview

Talk to your agents from the terminal

The Skydive CLI brings your agents to the terminal. Chat with them in a full TUI, send one-shot prompts from scripts, and manage agents, keys, and secrets from the command line.

It runs on macOS and Linux. On a Mac you can also use the [desktop app](/docs/desktop/mac); on Linux the CLI is how you give an agent access to your machine today.

## Install

```bash Binary (recommended) theme={null}
curl -fsSL https://skydive.sh | sh
```

```bash npm theme={null}
npm install -g skydive-cli
```

The install script drops a standalone `skydive` binary in `~/.local/bin`. No Node required, and `skydive update` upgrades it in place. Use npm if you would rather manage it alongside your other global packages.

Run it without installing anything:

```bash theme={null}
npx skydive-cli chat
```

### Beta channel

Beta builds carry the newest features ahead of stable, published on every change:

```bash Binary (recommended) theme={null}
SKYDIVE_CHANNEL=canary sh -c "$(curl -fsSL https://skydive.sh)"
```

```bash npm theme={null}
npm install -g skydive-cli@beta
```

Beta builds move fast and can break. Reinstall without the channel override to return to stable.

## Sign in

```bash theme={null}
skydive auth login --web
```

A browser page opens. Approve the sign-in and you're done: `skydive chat` works from this machine. Other sign-in options, including API keys for scripts, are covered in [Authentication](/docs/cli/authentication).

## Start chatting

```bash theme={null}
skydive chat
```

Pick an agent, pick a conversation, and talk. It is the same agent you reach in web chat and Slack, with the same memory.

> **[Chat TUI](/docs/cli/chat)**
>
> The interactive terminal experience: keybinds, images, themes, local shell.

> **[Scripting](/docs/cli/scripting)**
>
> One-shot prompts, JSON output, and pipelines with `chat -p`.

> **[Authentication](/docs/cli/authentication)**
>
> Web sign-in, API keys, and environment variables.

> **[Command reference](/docs/cli/reference)**
>
> Every command and flag.

## Commands at a glance

| Command | What it does |
| - | - |
| `skydive chat` | Open the interactive chat TUI, or `-p` for a one-shot prompt |
| `skydive auth` | Sign in, sign out, check auth state |
| `skydive agents` | List, inspect, and create agents |
| `skydive conversations` | List and read agent conversations |
| `skydive messages` | Re-fetch a reply by message id |
| `skydive workspace` | List workspaces or switch the active one |
| `skydive keys` | Manage API keys for an agent |
| `skydive secrets` | Manage an agent's secrets |
| `skydive portal` | Share this machine with an agent, manage access |
| `skydive sandbox` | Open a terminal in an agent's sandbox, or run a one-shot command there |
| `skydive import` | Import your Claude Code / Cursor / OpenCode setup into an agent |
| `skydive update` | Update the CLI in place |
| `skydive completion` | Install shell completions (bash, zsh, fish) |

Every command takes `--json` for machine-readable output and `--quiet` to print only the primary value. Run `skydive <command> --help` for details.


# Command reference
Source: https://getskydive.com/docs/cli/reference

Every skydive command and its flags

The full `skydive` command surface. Every command also answers `--help`, and all commands accept the global flags `--json`, `--quiet`, and `--api-url`.

## chat

Open the interactive [chat TUI](/docs/cli/chat), or send a [one-shot prompt](/docs/cli/scripting) with `-p`.

| Flag | Effect |
| - | - |
| `-p, --print [prompt]` | Non-interactive: send one prompt, print the reply, exit. Reads stdin when given no value. |
| `--agent <ref>` | Target agent by id, slug, or name. Without `-p`, skips the agent picker. |
| `--conversation <id>` | With `-p`: continue an existing conversation. |
| `--new` | Start a fresh conversation instead of resuming the last one. |
| `--share-machine` | Share this machine with the agent over Portal. |
| `--theme <id>` | Pin a colorscheme. Also settable via `SKYDIVE_THEME`. |
| `--no-notify` | Disable desktop notifications. |

## auth

Manage [authentication](/docs/cli/authentication).

| Subcommand | What it does |
| - | - |
| `auth login` | Store an API key (`--api-key sky_live_...`, or paste at the prompt) |
| `auth login --web` | Sign in for chat via the browser device flow |
| `auth status` | Show current authentication state |
| `auth logout` | Clear stored credentials (API key and chat session) |

## agents

| Subcommand | What it does |
| - | - |
| `agents list` | List agents in the active workspace (`--limit`, `--scope mine\|org`) |
| `agents get <id>` | Get agent details |
| `agents create` | Create a new agent (`--name`, `--model`) |

## conversations

| Subcommand | What it does |
| - | - |
| `conversations list` | List an agent's conversations (`--agent`, `--limit`) |
| `conversations show <id>` | Print a conversation transcript (`--recap` for the summary instead) |

## messages

| Subcommand | What it does |
| - | - |
| `messages get <message-id>` | Re-fetch an agent reply by id (the `messageId` from a prior `chat -p --json`) |

## workspace

| Subcommand | What it does |
| - | - |
| `workspace list` | List workspaces on your account |
| `workspace switch [workspace]` | Switch the active workspace by slug, name, or id. No argument picks interactively. |

## keys

API keys for an agent. All subcommands take `--agent-id <id>`.

| Subcommand | What it does |
| - | - |
| `keys list` | List API keys |
| `keys create <name>` | Create a key |
| `keys revoke <id>` | Revoke a key |

## secrets

An agent's [secrets](/docs/integrations/secrets). All subcommands take `--agent-id <id>`. Values are never shown after they are set.

| Subcommand | What it does |
| - | - |
| `secrets list` | List secret names |
| `secrets set <key> [value]` | Set a secret. Reads the value from stdin when omitted. |
| `secrets rm <key>` | Remove a secret |

## portal

Share this machine with agents. See [Reaching your computer](/docs/integrations/portal).

| Subcommand | What it does |
| - | - |
| `portal open` | Open Portal headless, in the foreground (`--agent` to grant one immediately, `--cwd` to set the working directory, `ctrl+c` to close) |
| `portal grant` | Grant an agent access to this machine |
| `portal revoke` | Remove an agent's access |
| `portal status` | Show machines, connection state, and granted agents |

## fs

Work with files on an agent's sandbox from your own machine.

| Subcommand | What it does |
| - | - |
| `fs edit <path>` | Open a remote file in your local editor (`$VISUAL`, then `$EDITOR`, then `vi`), and write your changes back when you save and close. A path that doesn't exist yet is created on save. Pass `--editor <cmd>` to override. |

```bash theme={null}
skydive fs edit --agent grace /tmp/notes.md
EDITOR=nano skydive fs edit ~/.self/soul.md
```

File access requires the same edit permission as `skydive sandbox`. The agent name or id can be provided with `--agent`.

## sandbox

Open a live terminal in an agent's sandbox, or run a one-shot command there.

| Usage | What it does |
| - | - |
| `sandbox` | Open an interactive sandbox terminal (needs a TTY) |
| `sandbox -- <cmd>` | Run one command in the sandbox and print its output |

## import

Import your Claude Code, Cursor, or OpenCode setup into an agent. Opens a machine-shared chat so the agent can read your local config.

## update

Update the CLI to the latest release in place.

| Flag | Effect |
| - | - |
| `--check` | Only report whether an update is available, don't install |
| `--channel <name>` | Update from a specific release channel |
| `--force` | Reinstall even when already current |

## config

View and change CLI preferences without editing `config.json` by hand.

```bash theme={null}
skydive config                       # list all preferences with their current values
skydive config get updateCheck       # print one key
skydive config set updateCheck false # change a preference
skydive config unset updateCheck     # revert to the built-in default
skydive config path                  # print the config.json location
```

Available preferences:

| Key | Type | Default | Effect |
| - | - | - | - |
| `updateCheck` | boolean | `true` | Check for CLI updates on startup |
| `shareMachineDefault` | boolean | `false` | Grant the agent Portal access automatically when you open chat |

`config` only exposes safe, hand-editable preferences. It cannot read or write credentials or flow-managed state.

The file itself lives at the platform-conventional path:

| Platform | Location |
| - | - |
| macOS | `~/Library/Preferences/skydive/config.json` |
| Linux | `~/.config/skydive/config.json` (or `$XDG_CONFIG_HOME/skydive/`) |
| Windows | `%APPDATA%\skydive\config.json` |

`skydive config path` prints the resolved path on your machine.

## logs

The CLI records local diagnostics by default. It keeps the latest 20 sessions.
Nothing uploads automatically. If you run into a bug or unexpected behavior,
export the latest session. The CLI can open a prefilled issue in the public
[Skydive CLI issue tracker](https://github.com/Create-Inc/skydive-cli), or you
can open the tracker and file the issue yourself. Attach the exported archive
to help us diagnose what happened.

| Subcommand | What it does |
| - | - |
| `logs list` | List recent diagnostic sessions |
| `logs path [session]` | Print a session directory. Defaults to the latest session. |
| `logs export [session]` | Create a `.tar.gz` archive and offer to open a prefilled GitHub issue. Defaults to the latest session. Use `--output` to choose the archive path. |

Automatic logs include request timing, TUI state changes, render performance,
and process metadata. Command values and URL query values are removed.

```bash theme={null}
skydive logs
skydive logs path
skydive logs export
```

## completion

Shell completions for bash, zsh, and fish.

| Subcommand | What it does |
| - | - |
| `completion install [shell]` | Install completions into your shell's startup file (auto-detects `$SHELL`) |
| `completion [shell]` | Print the completion script without installing |

## Environment variables

| Variable | Effect |
| - | - |
| `SKYDIVE_API_KEY` | Management API key, instead of `auth login` |
| `SKYDIVE_SESSION_TOKEN` | Chat session for non-interactive use |
| `SKYDIVE_THEME` | Pin a TUI colorscheme |
| `SKYDIVE_CHANNEL` | Release channel for the install script (`canary` for beta builds) |
| `SKYDIVE_API_URL` | Override the API base URL |
| `SKYDIVE_APP_URL` | Override the app base URL |
| `SKYDIVE_BUN_PATH` | Point chat at an existing Bun binary |
| `SKYDIVE_PROFILE` | Write a full profile to the working directory (`1`) or a specific path. Full profiles add CPU data. Set to `0` to disable all session diagnostics. |


# Scripting
Source: https://getskydive.com/docs/cli/scripting

One-shot prompts, JSON output, and pipelines

Use `chat -p` to send one prompt, print the reply, and exit. It never opens the TUI, runs under plain Node, and works without a TTY, so it fits scripts, pipes, and CI.

## One-shot prompts

```bash theme={null}
skydive chat -p "summarize my open PRs" --agent grace
```

Read the prompt from stdin by giving `-p` no value:

```bash theme={null}
echo "what changed today?" | skydive chat -p --agent grace
```

`--agent` takes an id, slug, or name. It is optional when your account has exactly one agent.

## JSON output

Add `--json` for structured output you can parse:

```bash theme={null}
skydive chat -p "list overdue invoices" --agent grace --json | jq .
```

The JSON includes the reply and a `messageId`. Re-fetch that reply later with [`skydive messages get`](/docs/cli/reference#messages).

## Continuing a conversation

Each `-p` call starts a new conversation by default. To keep context across calls, pass the conversation id back in:

```bash theme={null}
skydive chat -p "and which of those are over 30 days?" \
  --agent grace --conversation <conversation-id>
```

Get the id from the previous call's `--json` output.

## Auth for non-interactive use

Print mode never opens an interactive sign-in. In CI or scripts, authenticate ahead of time:

* Run `skydive auth login --web` once on the machine, or
* Set the `SKYDIVE_SESSION_TOKEN` environment variable.

See [Authentication](/docs/cli/authentication).

## Sharing your machine for a run

With `-p`, the `--share-machine` flag grants the target agent access to this machine for that run, so it can execute local commands as part of the task:

```bash theme={null}
skydive chat -p "run the test suite and summarize failures" \
  --agent grace --share-machine
```

## Global flags

Every `skydive` command accepts:

| Flag | Effect |
| - | - |
| `--json` | Output as JSON |
| `--quiet` | Print only the primary value |
| `--api-url <url>` | Override the API base URL (local development) |

`--quiet` pairs well with command substitution:

```bash theme={null}
AGENT_ID=$(skydive agents list --json | jq -r '.[0].id')
```


# Mac
Source: https://getskydive.com/docs/desktop/mac

Your agents on your Mac, working with the files and apps that live there

Skydive Desktop puts your agents on your Mac, so they can use the files, apps, and context that only exist on your machine.

That is the part a browser tab cannot do. Your agent reads what it needs from your Mac, pushes the work up to its own computer in the cloud, and keeps going from there. You can close the lid and it carries on.

Everything else is the same as the web: same agents, same conversations, same memory.

> [!NOTE]
>
> **macOS only for now.** Windows and Linux are coming. Until then, the [Skydive CLI](/docs/cli/overview) gives you the same agents in a terminal on Linux, and [`skydive portal`](/docs/cli/reference#portal) gives them access to the machine you are on. See [Not on a Mac](#not-on-a-mac).

## Install

**1. Download the app**

[Download Skydive Desktop](https://www.skydive.com/desktop), open the disk image, and drag Skydive into Applications.

**2. Sign in once**

There is no separate desktop account. Open the app, sign in to Skydive, and your agents and history are already there.

**3. Turn on access to this Mac**

Local access is off until you say otherwise. See [Letting an agent use your Mac](#letting-an-agent-use-your-mac).

| | |
| - | - |
| **Requires** | macOS 10.15 or later |
| **Chips** | Apple Silicon and Intel |
| **Updates** | Automatic, in the background |

The app also offers itself when you need it. Ask an agent in web chat for something that lives on your machine and a card appears with a download button.

## What the desktop app adds

> **[Your files and apps](/docs/integrations/portal)**
>
> Agents work with what is on your disk and the tools you have installed, not just what you upload.

> **Keeps running**
>
> Closing the window tucks the app into the menu bar. Your Mac stays reachable and work keeps moving.

> **Native saves**
>
> Files an agent produces save through the normal macOS save dialog.

> **Sign-ins that work**
>
> Connecting Google, Slack, and other services opens your real browser, so logins that block embedded windows still work.

## The loop

The point of the app is the handoff between two computers.

1. **Your Mac has the context.** A repo you have checked out, a dev server on `localhost`, a design file, a tool that only runs locally.
2. **The agent reaches it, with your permission.** You approve one agent at a time, and you can revoke it whenever you want.
3. **The work moves to the cloud.** The agent pushes what it collected up to [its own computer](/docs/capabilities/browser), where it has a browser, a shell, and no dependency on your laptop being awake.
4. **You get the result anywhere.** Web, Slack, email, or a text, because it is the same conversation on every channel.

## Using the app

The window is web chat, so everything in [Web chat](/docs/channels/web) applies. This page covers the parts that are only on the Mac.

### The menu bar

Skydive lives in the menu bar. Closing the window does not quit it, and neither does <kbd>Cmd</kbd> + <kbd>Q</kbd>. Both tuck it away so your Mac stays reachable and agents keep working.

| Item | What it does |
| - | - |
| **Status** | Whether this computer is reachable, and under what name |
| **Allow agents to use this computer** | The switch for local access, off by default |
| **Reload** | Bring the window back to Skydive |
| **Quit** | Actually exit, and drop the connection |

The status line tells you where you stand:

| It says | It means |
| - | - |
| `Open the app and sign in` | No session yet |
| `Off` | Local access is switched off, agents cannot reach this Mac |
| `Connecting…` | Coming online |
| `Connected` and your machine's name | Reachable |
| `Can't connect` and a reason | Something is wrong, and this says what |

Skydive starts at login, so your Mac comes online without you opening anything.

### Letting an agent use your Mac

Two switches, and both are yours.

**1. Open the machine**

Turn on **Allow agents to use this computer** in the menu bar. Nothing is reachable until you do. This is machine-level: it makes the Mac available, it does not grant anyone anything.

**2. Approve one agent**

An agent that wants local access asks for it, and a card appears in your chat. Approve it, and that one agent can reach this Mac. Nobody else can.

Revoke an agent whenever you want. Turning the menu bar switch off cuts access for every agent at once.

What an agent can actually do once approved, and how to ask for it, is in [Reaching your computer](/docs/integrations/portal).

> [!NOTE]
>
> Screenshots need one extra permission. The first time you turn on access, macOS asks for Screen Recording. Without it, an agent's screenshots come back blank.

### Continuing from the browser

When you are in web chat and an agent needs something local, it offers to move. A card appears with **Open Skydive**, the app comes forward, and the same conversation picks up where it left off. Approve the access request there and the agent starts.

If the app is not installed, the card offers the download instead.

### Updates

Skydive updates itself. It checks when it launches and every six hours after that, downloads a new version quietly, and applies it the next time the app starts.

To check by hand, open the **Skydive** menu and choose **Check for Updates…**. If one is ready, it offers to restart and apply it.

### Window and keyboard

| Keys | Action |
| - | - |
| <kbd>Cmd</kbd> + <kbd>+</kbd> / <kbd>-</kbd> | Zoom the interface in or out |
| <kbd>Cmd</kbd> + <kbd>0</kbd> | Back to actual size |
| <kbd>Cmd</kbd> + <kbd>R</kbd> | Reload |
| <kbd>Cmd</kbd> + <kbd>Opt</kbd> + <kbd>I</kbd> | Developer tools |
| Two-finger swipe | Back and forward, the same as Safari |

Links to other sites open in your normal browser. Skydive's own pages stay in the app. The window remembers its size and position.

### Troubleshooting

**The menu bar says `Off`.** Local access is switched off. Turn on **Allow agents to use this computer**.

**An agent says it cannot reach your machine.** Check the menu bar status first. If it reads `Connected`, the agent has no grant yet. Ask it to request access again, or approve the card in chat.

**Screenshots come back blank.** macOS Screen Recording is not granted. Open System Settings, then Privacy & Security, then Screen Recording, and allow Skydive.

**A sign-in page will not load in the app.** Use **Reload** from the menu bar to return the window to Skydive, then try connecting the service again.

> **[Portal](/docs/integrations/portal)**
>
> What an agent can actually do on your machine once you approve it.

## Not on a Mac

Windows and Linux builds are on the way. In the meantime the [Skydive CLI](/docs/cli/overview) covers the same ground from a terminal on Linux or macOS:

```bash theme={null}
curl -fsSL https://www.skydive.com/api/v1/cli/install.sh | sh
skydive auth login --web
skydive chat
```

That gives you the full [chat experience](/docs/cli/chat) in your terminal. For local access, press <kbd>ctrl</kbd> + <kbd>t</kbd> in chat to let an agent run commands on the machine you are on, or use [`skydive portal`](/docs/cli/reference#portal) to open it headlessly. It is the same permission model as the desktop app: default deny, one agent at a time, revocable.


# FAQ
Source: https://getskydive.com/docs/get-started/faq

Quick answers to common questions about Skydive

> **What is a Skydive agent, exactly?**
>
> A custom AI agent with its own cloud sandbox, persistent memory, connected services, and identity. You talk to it in plain language through web chat, Slack, email, or text. It does real work: it runs code, browses the web, calls APIs, and produces files.

> **Do I need to know how to code?**
>
> No. Most people build and run agents entirely through conversation. Agents write and run code inside their own sandbox. You never have to.

> **Where does my agent run?**
>
> Each agent runs in its own isolated cloud sandbox: a dedicated Linux machine provisioned on demand. Nothing runs on your computer. One agent's sandbox is isolated from another's.

> **How does the agent remember things?**
>
> Agents keep durable, file-based memory that persists across every conversation. Memory is agent-level: what one teammate teaches it is available to everyone who talks to it. Memory is stored as plain files in the agent's [git repository](/docs/agents/repo), so you can read, edit, or delete entries directly. See [Memory](/docs/agents/memory).

> **Is my data and are my credentials safe?**
>
> Skydive stores connected accounts and API keys securely and injects them at the network layer through a proxy. The agent acts on your behalf without seeing raw tokens in its prompts or logs. Memory lives in the agent's own repository, scoped to your workspace. Treat agent memory like a shared team wiki, not a vault for secrets. See [Permissioning](/docs/integrations/permissions).

> **Can my whole team use the same agent?**
>
> Yes. Agents live in a shared workspace. Invite teammates, and everyone can talk to the same agent across their own channels. See [Teams](/docs/workspaces/teams).

> **What channels can I reach my agent on?**
>
> Web chat, Slack, email, and SMS/iMessage. The agent keeps one identity and one memory across all of them. See [Channels](/docs/channels/overview).

> **Can an agent run on its own schedule?**
>
> Yes. Schedule an agent to run at a future time or on a recurring cadence: daily summaries, weekly reports, periodic checks. See [Routines](/docs/agents/routines).

> **How am I billed?**
>
> Plans are subscription-based. Paid plans include usage credit equal to the plan fee for the period (for example, Starter is \$20/month with \$20 of credit). Agents spend that credit balance on model usage, compute, and metered services. **Model usage is at cost: we do not mark it up.** Track it on the [Usage](/docs/account/usage) dashboard. See [Pricing](/docs/account/pricing).

> **Can I run more than one agent?**
>
> Yes. Most teams run several focused agents (one per job) rather than a single generalist. Starter, Team, and Enterprise all include unlimited agents; seats (workspace members) are what scale by plan. See [Creating agents](/docs/agents/creating-agents) and [Pricing](/docs/account/pricing).

> **What happens to an agent's work when its sandbox restarts?**
>
> Sandboxes are ephemeral. Agents save anything important (memory, shared workspace files, connected repositories) to durable storage automatically. Throwaway scratch work is not kept. The agent's configuration, its persona, tools, MCP connections, and memory files all live in its [git repository](/docs/agents/repo) and survive restarts.


# Quickstart
Source: https://getskydive.com/docs/get-started/quickstart

Build and talk to your first agent

This guide takes you from a new account to a working agent. It takes about five minutes.

## 1. Create your account

Go to [skydive.com](https://skydive.com/signup) and sign up with Google or Slack. Skydive creates a personal **workspace** for you on first load. Your agents and teammates live in the workspace.

> [!TIP]
>
> Sign up with Slack or Google to make that channel ready to connect to your agents later.

## 2. Create an agent

First-run setup has three steps:

1. **Name your agent and give it a goal.** Pick a character and a name. Add a one-line goal that says what the agent is for. The goal seeds the agent's instructions. You can refine it anytime.
2. **Connect its services.** Skydive suggests services based on the goal you typed, such as GitHub or a Google account. Connect them now, or skip and do it later.
3. **Pick its channels.** Add the agent to Slack, or give it its own email inbox.

Skydive then provisions the agent's sandbox: its own cloud machine, a fresh memory store, and its baseline tools. This takes a few seconds.

> [!NOTE]
>
> To create another agent later, click **New agent** on your dashboard, then name it and give it a goal.

## 3. Say hello

Open the agent and send your first message. Describe the outcome you want, not the steps.

```text theme={null}
Hey! Can you research the top 3 competitors to Notion and
put together a short comparison table?
```

The agent plans, works in its sandbox, and streams progress back to you. It asks when it needs a decision or a credential.

## 4. Connect a service (optional)

When a task needs access to a service, the agent prompts you to connect it. Authorization takes one click. Skydive stores the credential securely. The agent uses it through a proxy and never sees the raw token.

See [Connecting services](/docs/integrations/oauth) for the full flow.

## 5. Put it where you work

Once the agent is useful, add it to your channels:

> **[Add it to Slack](/docs/channels/slack)**
>
> DM your agent or @-mention it in channels.

> **[Email it](/docs/channels/email)**
>
> Give your agent its own inbox.

> **[Text it](/docs/channels/text)**
>
> Reach it over SMS or iMessage.

> **[Schedule it](/docs/agents/routines)**
>
> Run work on a cadence.

## Next steps

> **[Talking to your agent](/docs/get-started/talking-to-your-agent)**
>
> Get more out of every conversation.

> **[Give it a persona](/docs/agents/persona)**
>
> Shape how your agent thinks and writes.

> **[Invite your team](/docs/workspaces/teams)**
>
> Share the agent with teammates.

> **[Understand pricing](/docs/account/pricing)**
>
> Pick the right plan.


# Talking to your agent
Source: https://getskydive.com/docs/get-started/talking-to-your-agent

How to prompt, collaborate, and get great results

You get better results when you describe outcomes instead of steps. This page shows how to prompt an agent well.

## Describe outcomes, not steps

Describe the result you want and the context around it. The agent plans the steps itself.

```text Good theme={null}
Put together a competitive teardown of our top 3 rivals: pricing,
positioning, and what they ship weekly. A short doc I can share with
the team is perfect.
```

```text Less effective theme={null}
Open a browser. Go to each competitor site. Copy the pricing. Paste
it into a doc.
```

The first prompt gives the agent a goal and a finish line. The second removes its judgment.

## Give it context

The agent knows what you tell it plus what it remembers. Include the things only you know:

* Who the work is for and how it will be used
* Constraints: deadline, tone, format, must-include, must-avoid
* Where to find inputs ("the spec is in the file I just dropped in")

## Answer its questions

The agent asks before it guesses on anything consequential: a destructive action, a spend, an ambiguous requirement. Answer the question. The result is better than if it had assumed.

## Work in passes

Treat the first output as a draft, then refine it:

```text theme={null}
Great start. Tighten the intro, drop the third section, and make the
tone a notch more direct.
```

## Hand off recurring work

Recurring and tedious work is where agents help most:

* "Every weekday at 9am, summarize what changed in our shared inbox." → a [routine](/docs/agents/routines)
* "Whenever I forward you a contract, pull out the key dates and obligations."
* "Keep our launch checklist current as we close items in Slack."

## Todos

On longer work, the agent keeps a live plan in the conversation. You see it as a checklist: what it is doing, what is next, what is done. You do not fill it in. You can glance at it to see where things stand.

It is not a project tracker. It exists for the current run and disappears when the work is finished. Recurring work belongs in a [routine](/docs/agents/routines).

## Teach it once

When you correct the agent or state a durable preference, it writes that to [memory](/docs/agents/memory). You don't repeat yourself next time, and your teammates get the same behavior.

```text theme={null}
Going forward, always use British spelling in anything customer-facing.
```

## Attach files and images

Drop files into the conversation: documents, spreadsheets, screenshots, audio. The agent reads, analyzes, transcribes, and transforms them, then returns files you can download.

> **[Shape its default behavior](/docs/agents/persona)**
>
> Set a persona and standing instructions once instead of repeating preferences in every chat.


# Troubleshooting errors
Source: https://getskydive.com/docs/get-started/troubleshooting-errors

What each error message means and what to do about it

When a run fails, the message you see is trying to tell you something specific. This page decodes every error message Skydive shows, in one place. If the message names a fix, follow it. Everything here works the same on every channel: the wording is identical in web chat, Slack, email, and text.

## What to try first

One retry is almost always safe. If a retry produces the same error, use the table below to find the exact message and follow its fix. Errors that say "try again in a moment" usually clear on their own within minutes.

## Run errors

These appear in the conversation where the run failed, and in run history.

| Message | What it means | What to do |
| - | - | - |
| **This conversation is too long for the model. Start a new thread to continue.** | The conversation has grown past what the model can hold at once. Nothing is broken. | Start a new thread. The agent's memory carries over, so it keeps the context that matters. |
| **Your agent hit its usage limit and stopped. It resumes automatically in a few minutes; you can also retry now.** | The agent hit its rate limit. This is temporary, not a billing problem. | Wait a few minutes and it picks back up on its own, or retry now if the work is urgent. |
| **Your workspace's model provider key doesn't have enough credits for this request. Add credits, choose another model, or contact a workspace admin.** | The workspace's own connected provider (a key your workspace added) ran out of credits. The failure came from your key, not Skydive's. | Add credits at the provider, switch the agent to a different model, or ask a workspace admin to top up the key. |
| **Your Claude subscription is refusing requests because it can't verify that model training is off. In Claude, open Settings > Privacy and turn off 'Help improve our AI models', then retry.** | The Claude account your workspace connected is refusing requests until training consent is verified off. The fix is in Claude, not Skydive. | In Claude: **Settings > Privacy**, turn off **Help improve our AI models**, then retry in Skydive. |
| **Your request was blocked by the AI provider's content filter. Try rephrasing your request and sending it again.** | The model provider's safety filter rejected the request. | Rephrase and resend. Rephrasing usually clears it. |
| **The model is temporarily unavailable. Please try again in a moment.** | The model provider is overloaded or unreachable right now. | Try again in a moment. This clears on its own. |
| **The model stopped responding before this request finished. Please try again.** | A generic provider failure: the model cut out mid-run without a more specific reason. | Retry once. If it repeats on the same request, the provider connection may be the problem; try a different model or contact support. |
| **The service that runs the agent's computer is temporarily unavailable. Please try again.** | The agent's sandbox couldn't be started at all. Nothing was lost. | Retry in a few minutes. |

## Plan and compute errors

| Message | What it means | What to do |
| - | - | - |
| **Your connected plan can't take this request right now. Wait for its usage window to reset, connect another account, or switch the model to run on Skydive.** | The BYOK plan the model runs on (your own Claude/OpenAI subscription) hit its usage cap. | Wait for the plan's window to reset, connect another account, or switch the agent to a model that runs on Skydive's credits. |
| **The agent's computer isn't responding. Nothing was lost.** | The agent's sandbox crashed or was killed mid-run. Completed work up to the failure point is saved. | Retry. If the same task keeps dying this way, the task may be too big for the agent's compute tier; ask for it in smaller steps or size the agent up. |

## Text message failures (iMessage/SMS)

On text, failure notices are shorter and written in the agent's voice. They decode to the same errors as above:

| Notice | It means |
| - | - |
| "this conversation has gotten too long for the model to keep replying in it" | Conversation too long. Start a new thread on web. |
| "you've hit the rate limit for now, i'll pick back up automatically in a few minutes, or you can retry now" | Usage limit hit. It self-resumes. |
| "your workspace's connected ai provider ran out of credits, so i can't take this right now" | Your workspace's provider key needs credits. |
| "the model's content filter blocked this one, try rephrasing your message and sending it again" | Content filter. Rephrase. |
| "the model is overloaded with traffic right now, try again in a bit" | Provider overload. Wait and retry. |
| "sorry, skydive is experiencing some strain so responses might be slow, i'll pick it up soon" | An infrastructure problem on our side (the agent's computer didn't start, died, or timed out). Nothing you did. No action needed; it picks back up. |
| "something interrupted my reply, try again in a bit" | An unclassified failure. Retry. |

> [!NOTE]
>
> In a group text, failure notices are suppressed entirely. If a run fails in a group thread, you'll see the agent simply stop without a notice.

## Email domain errors

These appear when connecting or verifying a custom email domain. Full setup steps are in [Email](/docs/channels/email#custom-domains).

| Message | What it means | What to do |
| - | - | - |
| **This domain is already registered** | Either another workspace owns the domain, a deleted workspace's registration is still held (support can reclaim it), or your domain's MX points at another mail provider (like Google Workspace) and needs the coexistence flow. | If your domain runs Google Workspace, follow the [coexistence steps](/docs/channels/email#google-workspace-coexistence). Otherwise contact support. |
| **Taken** (for an address you've never used) | The address was claimed by another workspace at some point. Addresses are permanent and never reissued. | Pick a different address. |
| **Domain limit exceeded** | Your workspace hit its custom-domain cap. | Retrying won't help; contact support to raise the cap. |
| **Provider error, try again** (when adding a domain) | A retry usually won't fix this: the common causes are a conflicting provider on your domain, an orphaned registration, or a quota cap. | Check your domain's MX records first, then contact support with the error text and your domain name. |
| **Verification stuck on "Verifying" with all records showing valid** | A rare authorization issue on our side. | Contact support with your domain name. |

## What to send when contacting support

For anything the table doesn't answer, email [support@skydive.com](mailto:support@skydive.com) with:

* The exact error text (copy it verbatim from the conversation or run history)
* Which channel it happened on (web, Slack, email, text)
* What the agent was doing when it failed

> **[Run history](/docs/workspaces/feed)**
>
> Revisit past runs and their errors in the feed.

> **[Usage and billing](/docs/account/usage)**
>
> Where credit and usage limits show up.


# Video tutorials
Source: https://getskydive.com/docs/get-started/video-tutorials

Watch Skydive tutorials on YouTube

Skydive walkthroughs on YouTube. One playlist powers this page: add a tutorial to that playlist and it shows up in the player with no docs edit. The pick list below is kept in sync twice a day.

## Watch

Play the series below. Open the playlist panel on the player to jump between videos without leaving the page.

[Skydive video tutorials](https://www.youtube-nocookie.com/embed/videoseries?list=PLck6pvHwS-WI)

## Pick a video

Jump straight to a title. Each link opens that video inside the same playlist.

| Video | Open |
| - | - |
| Welcome to Skydive | [Watch](https://www.youtube.com/watch?v=nd-6EzORoRE\&list=PLck6pvHwS-WI) |
| How to use Notion with Skydive | [Watch](https://www.youtube.com/watch?v=1urpLyP1CG8\&list=PLck6pvHwS-WI) |
| How to use Canva with Skydive | [Watch](https://www.youtube.com/watch?v=Uj5VDbQaFWM\&list=PLck6pvHwS-WI) |
| A tour of your Skydive Agent + Homepage | [Watch](https://www.youtube.com/watch?v=845wdghnt_U\&list=PLck6pvHwS-WI) |
| How to set a routine with your agent | [Watch](https://www.youtube.com/watch?v=htT_nyOmNxU\&list=PLck6pvHwS-WI) |

Full playlist: [New to Skydive? Start here](https://www.youtube.com/playlist?list=PLck6pvHwS-WI). Channel: [youtube.com/@getskydive](https://www.youtube.com/@getskydive).

## Related docs

> **[Quickstart](/docs/get-started/quickstart)**
>
> First agent in a few minutes.

> **[Slack](/docs/channels/slack)**
>
> Install Skydive in Slack.

> **[Notion](/docs/integrations/notion)**
>
> Connect Notion and keep pages flowing.

> **[Canva](/docs/integrations/canva)**
>
> Connect Canva for designs and brand kits.


# Welcome
Source: https://getskydive.com/docs/get-started/welcome

Build AI agents that do real work across your tools

Skydive lets anyone on your team build AI agents that do real work across your tools.

You talk to an agent in Slack, email, or iMessage, and each one gets its own Slack bot, inbox, and phone number. Or reach it on the web, in the desktop app, or from your terminal.

Every agent has its own computer in the cloud, with a browser, a terminal, and a file system. Give it access to the apps it needs, teach it the role, and it keeps working after everyone closes their laptop.

Creating an agent takes less than 30 seconds.

> **[Build your first agent](/docs/get-started/quickstart)**
>
> Zero to a working agent in a few minutes.

> **[Video tutorials](/docs/get-started/video-tutorials)**
>
> Watch it done end to end.

## What makes Skydive different

Make your own team of agents.

> **[Agents work where you already do](/docs/channels/overview)**
>
> The same agent in Slack, email, iMessage, web, desktop, and the terminal.
>
> One identity and one memory across all of them.

> **[Agents have their own computer](/docs/capabilities/browser)**
>
> A cloud machine with a browser, terminal, and file system.
>
> It fills out sites, runs code, builds apps, and moves between tools until the job is done.

> **[Agents work inside your tools](/docs/integrations/catalog)**
>
> Connect it to your real accounts in one click.
>
> Where there is no clean integration, it hands you its browser so you can log in.

> **[Agents run without you](/docs/agents/routines)**
>
> Ask it to watch an inbox, a metric, or a deploy, and it turns that into a routine that runs on a schedule or on an event.
>
> It runs in the cloud, so it keeps working while your laptop is closed.

> **[Agents work with other agents](/docs/capabilities/subagents)**
>
> Put several in a group chat and they share context and hand work off.
>
> You stay in the loop and can redirect at any point.

> **[Agents get better as you work](/docs/agents/memory)**
>
> It remembers your feedback, learns from corrections, and writes its own skills.
>
> Every agent is specialized rather than one generalist.

## How it works

**1. Create an agent**

Give it a name and a job. Skydive provisions its computer, memory, and identity.

**2. Tell it what you want**

Describe the outcome. It plans, acts, and reports back, and asks when it needs a decision or a credential.

**3. Connect what it needs**

Approve services as it asks for them. Credentials go through a proxy, so the agent never sees the raw token.

**4. Put it to work**

Add it to Slack, give it a routine, and share it with your team.

## What people build

You can make as many Skydive agents as you need. Make a new one for each role, as big or small as you need.

Support agents that answer in Intercom. Engineering agents that debug production and open PRs. Recruiting agents that schedule first calls. Marketing agents that generate creative and run campaigns.

> **[231 ways people use Skydive agents](https://www.skydive.com/blog/231-agent-use-cases)**
>
> Real jobs across engineering, support, sales, ops, and design.

> **[Don't stop at one agent](https://www.skydive.com/blog/bench)**
>
> Why a bench of focused agents beats one overloaded generalist.

> **[When agents become teammates](https://www.skydive.com/blog/teammates)**
>
> An engineering agent shipped three PRs before lunch on its first day.

> **[How to train your agent](https://www.skydive.com/blog/how-to-train-your-agent)**
>
> Turning a functional agent into one you trust.

## Keep going

> **[Agents](/docs/agents/overview)**
>
> What an agent is made of

> **[Talking to your agent](/docs/get-started/talking-to-your-agent)**
>
> How to prompt and collaborate

> **[Workspaces & teams](/docs/workspaces/overview)**
>
> Share agents with your team

> **[Plans & pricing](/docs/account/pricing)**
>
> Find the right plan

> **[FAQ](/docs/get-started/faq)**
>
> Quick answers

> **[Support](/docs/support)**
>
> Get help from the team


# Read AI
Source: https://getskydive.com/docs/integrations/account

AI meeting notes, summaries and transcripts

## Example agent tasks

```text theme={null}
Connect Read AI and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Read AI.
```

```text theme={null}
Use the agent to interact with Read AI directly from your conversation.
```

## Connecting

Connect Read AI over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Read AI.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Read AI through the MCP server.

## What you share

Connecting grants the agent access to Read AI on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Acuity Scheduling
Source: https://getskydive.com/docs/integrations/acuity-scheduling

Let your agent manage appointments and clients in Acuity Scheduling

Acuity Scheduling is an online appointment booking platform for managing availability, appointments, and client intake. Connect it and your agent can read your calendar, handle scheduling logic, and work with client records on your behalf.

## What the agent can do

> **Read appointments**
>
> Pull upcoming and past appointments across your calendars.

> **Manage availability**
>
> Check open slots and work with your scheduling rules.

> **Work with clients**
>
> Look up client records and intake form responses.

> **Handle appointment types**
>
> Read the appointment types and durations you've configured.

## Example agent tasks

```text theme={null}
Show me everything booked for tomorrow and flag any gaps I could fill.
```

```text theme={null}
List all clients who completed an intake form this week and what they submitted.
```

```text theme={null}
Every morning, give me a summary of today's appointments with client names and notes.
```

## What you share

During the connect flow you authorize what the agent can reach. It works only within what you grant. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Authorize Acuity Scheduling for the agent**

From the agent's connections, choose Acuity Scheduling and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[Acuity Scheduling developer docs](https://developers.acuityscheduling.com/reference)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Advanced Web Ranking
Source: https://getskydive.com/docs/integrations/advancedwebranking

SEO rank tracking and reporting

## Example agent tasks

```text theme={null}
Connect Advanced Web Ranking and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Advanced Web Ranking.
```

```text theme={null}
Use the agent to interact with Advanced Web Ranking directly from your conversation.
```

## Connecting

Connect Advanced Web Ranking over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Advanced Web Ranking.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Advanced Web Ranking through the MCP server.

## What you share

Connecting grants the agent access to Advanced Web Ranking on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Affinity
Source: https://getskydive.com/docs/integrations/affinity

Relationship intelligence CRM for dealmakers

Affinity is a relationship intelligence CRM for dealmakers. Connect its hosted MCP server and your agents can search people, companies, and opportunities, work list entries and fields, write notes, and manage reminders, all scoped to what your API key can access.

## What your agents can do

> **Search relationships**
>
> Search people, companies, and opportunities with structured filters or natural language, and pull enriched profiles and field data.

> **Work lists and fields**
>
> Read, create, and update list entries and field values across your pipeline lists, saved views, and dropdown options.

> **Capture notes and meetings**
>
> Create, update, and search notes, upload and download files, read meeting transcripts, and log interactions.

> **Map warm paths and reminders**
>
> Read relationship strengths and coworker connections, render warm intro graphs, and create or update reminders.

## Example agent tasks

```text theme={null}
Search Affinity for people we have not interacted with in over six months and list them by relationship strength.
```

```text theme={null}
Add this contact to our Founders list, create a note explaining the intro, and set a reminder to follow up next week.
```

```text theme={null}
Find the warmest intro path from our team to this company and draft the outreach note.
```

## Connecting

**Affinity MCP server**

Connect to Affinity's official hosted MCP server with an API key. Requests authenticate with a Bearer token, and tools stay within the API key owner's permissions.

**1. Create an API key**

You need an Affinity Scale, Advanced, or Enterprise plan. Create an API key under Settings > Manage Apps in Affinity.

**2. Add the connection**

From your agent's connections, choose Affinity and paste the API key when prompted.

**3. The agent gets to work**

Start with search\_persons or search\_opportunities, then work lists, notes, and reminders from there.

Write tools create and update real records in Affinity. Deleting or merging people and companies is destructive, so agents confirm with you first.

## What you share

Connecting grants your agents access to Affinity through its MCP server, scoped to the permissions of the API key you provide. They can read and search people, companies, opportunities, lists, notes, files, meetings, and reminders, and write the same record types. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

## Learn more

> **[Affinity MCP available tools](https://developer.affinity.co/pages/mcp/available-tools)**
>
> The full tool list with example prompts for each.

> **[MCP authentication](https://developer.affinity.co/pages/mcp/authentication)**
>
> How to create and use your Affinity API key.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Agentcard
Source: https://getskydive.com/docs/integrations/agentcard

Let your agent create and manage virtual debit cards

Agentcard issues virtual debit cards for AI agents. Connect it and your agent can create cards, fund them, track transactions, and shop at linked merchants on your behalf.

> [!WARNING]
>
> Card creation, purchases, and withdrawals move real money and are irreversible. Your agent will always confirm the exact amount and merchant before spending.

## What the agent can do

> **Create virtual cards**
>
> Issue single-use or multi-use virtual debit cards funded from your balance.

> **Shop at linked merchants**
>
> Place purchases at connected merchants like DoorDash in natural language.

> **Track transactions**
>
> Pull transaction history, balances, and card status across your wallet.

> **Manage your wallet**
>
> Add funds, withdraw cash, set card limits, pause or close cards.

## Example agent tasks

```text theme={null}
Create a $50 virtual card for this software purchase and close it after.
```

```text theme={null}
What's my current balance and how many cards do I have active?
```

```text theme={null}
Order dinner from DoorDash tonight, something under $30.
```

## What you share

Your Agentcard account must be KYC-verified and funded before the agent can create cards. The agent operates within what your account allows and confirms amounts before any spending action. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Add the Agentcard connection**

From the agent's connections, choose Agentcard. The platform connects you through Agentcard's MCP server.

**2. Sign in to Agentcard**

A connect card appears. Sign in to your Agentcard account and approve.

**3. Complete KYC if needed**

Card creation requires identity verification. If your account isn't verified yet, the agent will walk you through it.

**4. The agent gets to work**

It can now create cards, track transactions, and shop at linked merchants.

## Learn more

> **[Agentcard](https://agentcard.sh)**
>
> Learn more about virtual cards for agents.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Agiflow
Source: https://getskydive.com/docs/integrations/agiflow

AI-native project board for solo builders and small teams

## Example agent tasks

```text theme={null}
Connect Agiflow and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Agiflow.
```

```text theme={null}
Use the agent to interact with Agiflow directly from your conversation.
```

## Connecting

Connect Agiflow over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Agiflow.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Agiflow through the MCP server.

## What you share

Connecting grants the agent access to Agiflow on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Agility CMS
Source: https://getskydive.com/docs/integrations/agilitycms

Azure-hosted headless CMS for managing multiple sites

## Example agent tasks

```text theme={null}
Connect Agility CMS and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Agility CMS.
```

```text theme={null}
Use the agent to interact with Agility CMS directly from your conversation.
```

## Connecting

Connect Agility CMS over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Agility CMS.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Agility CMS through the MCP server.

## What you share

Connecting grants the agent access to Agility CMS on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Ahrefs
Source: https://getskydive.com/docs/integrations/ahrefs

SEO tools for backlinks and keyword research

## Example agent tasks

```text theme={null}
Connect Ahrefs and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Ahrefs.
```

```text theme={null}
Use the agent to interact with Ahrefs directly from your conversation.
```

## Connecting

Connect Ahrefs over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Ahrefs.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Ahrefs through the MCP server.

## What you share

Connecting grants the agent access to Ahrefs on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Airbyte Agent
Source: https://getskydive.com/docs/integrations/airbyte

Data pipeline and ETL connector platform

## Example agent tasks

```text theme={null}
Connect Airbyte Agent and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Airbyte Agent.
```

```text theme={null}
Use the agent to interact with Airbyte Agent directly from your conversation.
```

## Connecting

Connect Airbyte Agent over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Airbyte Agent.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Airbyte Agent through the MCP server.

## What you share

Connecting grants the agent access to Airbyte Agent on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Airtable
Source: https://getskydive.com/docs/integrations/airtable

Let your agent read and update your Airtable bases

Airtable is a flexible database that looks like a spreadsheet. Connect it and your agent can read your tables, filter records, and create or update rows.

## Two ways to connect

Airtable now runs a hosted MCP server. For most people, that is the easier path: one connection, broad access, no need to pick individual bases.

* **Airtable MCP (recommended).** Connect Airtable's MCP server once and the agent can read and write across your bases, create tables and automations. See [Connect with MCP](#connect-with-mcp).
* **Airtable OAuth.** The original sign-in integration, scoped to what you grant. See [Connect with OAuth](#connect-with-oauth).

## What the agent can do

> **Read records**
>
> Pull rows from a table, filtered and sorted how you need.

> **Create records**
>
> Add new rows to a table.

> **Update records**
>
> Change field values on existing rows.

> **Find a base**
>
> Browse your bases and tables.

## Example agent tasks

```text theme={null}
When a lead comes in by email, add them to my CRM base with the source and a follow-up date.
```

```text theme={null}
Every Friday, tell me which projects are behind and nudge the owners.
```

```text theme={null}
Keep my content calendar in sync: when I approve a draft, flip its status and set the publish date.
```

## Connect with MCP

**1. Add the Airtable connection**

From the agent's connections, choose Airtable. The platform connects you through Airtable's MCP server.

**2. Sign in to Airtable**

A connect card appears. Sign in to your Airtable account and approve.

**3. The agent gets to work**

It can now read and write across your bases, create tables and automations through the MCP server.

> [!NOTE]
>
> MCP is the first-class connection when both options are available. See [Tools & MCP](/docs/capabilities/tools-and-mcp).

## Connect with OAuth

## What you share

During the connect flow you authorize what the agent can reach. It works only within what you grant. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

**1. Authorize Airtable for the agent**

From the agent's connections, choose Airtable and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[Airtable API docs](https://airtable.com/developers/web/api/introduction)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Airwallex AgentOS
Source: https://getskydive.com/docs/integrations/airwallex

Global business payments and financial infrastructure

## Example agent tasks

```text theme={null}
Connect Airwallex AgentOS and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Airwallex AgentOS.
```

```text theme={null}
Use the agent to interact with Airwallex AgentOS directly from your conversation.
```

## Connecting

Connect Airwallex AgentOS over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Airwallex AgentOS.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Airwallex AgentOS through the MCP server.

## What you share

Connecting grants the agent access to Airwallex AgentOS on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Alpaca
Source: https://getskydive.com/docs/integrations/alpaca

Commission-free stock and crypto trading API

## Example agent tasks

```text theme={null}
Connect Alpaca and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Alpaca.
```

```text theme={null}
Use the agent to interact with Alpaca directly from your conversation.
```

## Connecting

Connect Alpaca over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Alpaca.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Alpaca through the MCP server.

## What you share

Connecting grants the agent access to Alpaca on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Alpic
Source: https://getskydive.com/docs/integrations/alpic

Hosting and deployment platform for MCP servers

## Example agent tasks

```text theme={null}
Connect Alpic and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Alpic.
```

```text theme={null}
Use the agent to interact with Alpic directly from your conversation.
```

## Connecting

Connect Alpic over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Alpic.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Alpic through the MCP server.

## What you share

Connecting grants the agent access to Alpic on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Ambient
Source: https://getskydive.com/docs/integrations/ambient

Shared team brain for meetings, Slack, and email.

Ambient keeps a searchable record of your team's meetings, Slack conversations, and email threads. Your agents can pull meeting summaries, transcripts, action items, and project context on demand.

## What the agent can do

> **Search meeting summaries**
>
> Query past meetings by topic, participant, or date range and get a concise summary of what was discussed.

> **Retrieve transcripts**
>
> Pull the full transcript for any recorded meeting so the agent can answer follow-up questions or extract specific details.

> **Surface action items**
>
> Find open action items from any meeting and bring them into whatever workflow the agent is running.

> **Access project context**
>
> Read project-level meeting history so your agents have the background they need to work on a task without you having to re-explain it.

## Example agent tasks

```text theme={null}
Summarize what was discussed in last week's product review meeting.
```

```text theme={null}
Find all action items assigned to me from meetings in the past month.
```

```text theme={null}
Pull the transcript from yesterday's customer call and list the key concerns raised.
```

## What you share

Connecting grants your agents read-only access to your Ambient workspace. They can read meetings, summaries, transcripts, and projects but cannot create or modify anything. See [Permissioning](/docs/integrations/permissions).

## Connecting

**MCP server**

Connect Ambient over MCP for direct access to meeting summaries, transcripts, projects, and insights. This is the recommended path once your Ambient account has API access enabled.

**1. Enable API access**

Ambient API access is not on by default. Open Account Settings in your Ambient workspace and look for the Ambient API section. If it is not there, email [support@ambient.us](mailto:support@ambient.us) to request access.

**2. Add the connection**

From the agent's tools, add Ambient. A connect card appears.

**3. Authorize**

Approve Ambient when prompted. The platform handles the OAuth handshake.

**4. The agent gets to work**

Your agents can now search meetings, retrieve transcripts, and pull project context from your Ambient workspace.

**API key**

Connect Ambient REST with a Bearer API key from Account Settings. Use this if your Ambient account does not yet have MCP OAuth available.

**1. Enable API access**

Ambient API access is not on by default. Open Account Settings in your Ambient workspace and look for the Ambient API section. If it is not there, email [support@ambient.us](mailto:support@ambient.us) to request access.

**2. Create an API key**

In Account Settings, open the Ambient API section and create a new API key. Copy it right away; Ambient only shows it once.

**3. Add the Ambient connection**

From the agent's connections, choose Ambient. Paste your API key when prompted.

**4. The agent gets to work**

Your agents can now search meetings, retrieve transcripts, and pull project context from your Ambient workspace.

## Learn more

> **[Ambient help center](https://app.ambient.us/help)**
>
> Documentation and guides for Ambient.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Amplemarket
Source: https://getskydive.com/docs/integrations/amplemarket

AI-powered sales engagement and prospecting platform.

Amplemarket is an AI-powered sales engagement and prospecting platform. Connect it and your agents can search leads, enrich contacts and companies, manage sequences and mailboxes, and work call and email data without leaving their workflow.

## What your agents can do

> **Search and manage leads**
>
> Find prospects, add them to lead lists, and keep list membership current.

> **Enrich contacts and companies**
>
> Pull people and company enrichment data, including emails and firmographics.

> **Work sequences and mailboxes**
>
> Read sequence activity and mailbox state so your agents can report on outreach.

> **Review calls and tasks**
>
> Retrieve call recordings and dispositions, and keep an eye on open tasks.

## Example agent tasks

```text theme={null}
Find contacts at companies matching our ideal customer profile and enrich them with verified emails.
```

```text theme={null}
Add this week's inbound signups to our outbound lead list and report who is already in a sequence.
```

```text theme={null}
Summarize call dispositions from the past month and flag deals that went quiet.
```

## What you share

Connecting grants your agents access to your Amplemarket workspace through its MCP server or REST API. They can read and, where the tool supports it, act on leads, contacts, enrichment, sequences, mailboxes, calls, and tasks. See [Permissioning](/docs/integrations/permissions).

## Connecting

**MCP server**

Connect Amplemarket over MCP for direct access to its prospecting and sales tools. Authorization uses OAuth with your Amplemarket account.

**1. Add the connection**

From your agent's tools, add Amplemarket. A connect card appears.

**2. Authorize**

Sign in to Amplemarket when prompted and approve the connection. The platform handles the OAuth handshake.

**3. The agent gets to work**

Your agents can now search leads, enrich contacts, and work sequences from your Amplemarket workspace.

**API key**

Connect Amplemarket through its REST API with a Bearer API key.

**1. Create an API key**

In the Amplemarket dashboard go to Settings > API and click + New API Key to mint one. Copy it right away.

**2. Add the Amplemarket connection**

From your agent's connections, choose Amplemarket and paste your API key when prompted.

**3. The agent gets to work**

Your agents can now call the Amplemarket REST API, from lead lists to email validation.

## Learn more

> **[Amplemarket API reference](https://docs.amplemarket.com/api-reference)**
>
> Endpoint reference for the Amplemarket REST API.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Amplitude
Source: https://getskydive.com/docs/integrations/amplitude

Product analytics and user behavior tracking

## Example agent tasks

```text theme={null}
Connect Amplitude and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Amplitude.
```

```text theme={null}
Use the agent to interact with Amplitude directly from your conversation.
```

## Connecting

Connect Amplitude over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Amplitude.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Amplitude through the MCP server.

## What you share

Connecting grants the agent access to Amplitude on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Anything
Source: https://getskydive.com/docs/integrations/anything

Let your agent build, publish, and manage AI-generated apps on Anything

Anything is an AI-powered app builder. Connect it with an API key and your agent can create projects, generate app code, publish finished apps, and manage your organization.

## What the agent can do

> **Create projects**
>
> Spin up new app projects with a prompt and publish them when they're ready.

> **Generate app code**
>
> Drive iterative code generation on an existing project, refining the app through conversation.

> **Publish and ship**
>
> Publish a project live or use the ship command to go from idea to deployed app in one step.

> **Manage organizations**
>
> List organizations, check project status, and manage secrets across environments.

## Example agent tasks

```text theme={null}
Create a new Anything app that tracks my weekly goals and publish it.
```

```text theme={null}
List my Anything projects and tell me which ones are currently published.
```

```text theme={null}
Generate a simple expense tracker app on Anything and deploy it.
```

## What you share

Connecting grants the agent access to your Anything account using an API key scoped to your user. It can create and publish projects within your organizations. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Get an API key**

Go to [anything.com/dashboard/settings](https://anything.com/dashboard/settings) under API Keys. Click Create key, then copy it. The key is shown only once.

**2. Add the Anything connection**

From the agent's connections, choose Anything. Paste your API key when prompted.

**3. The agent gets to work**

It can now create projects, generate apps, and publish them to Anything.

## Learn more

> **[Anything docs](https://anything.com/docs)**
>
> Full documentation for the Anything platform.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Apify
Source: https://getskydive.com/docs/integrations/apify

Web scraping and browser automation platform

## Example agent tasks

```text theme={null}
Connect Apify and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Apify.
```

```text theme={null}
Use the agent to interact with Apify directly from your conversation.
```

## Connecting

Connect Apify over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Apify.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Apify through the MCP server.

## What you share

Connecting grants the agent access to Apify on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# APIs
Source: https://getskydive.com/docs/integrations/apis

Connect anything with an API key or token

When a service offers keys or tokens but not OAuth or MCP, paste the key once. Skydive stores it as a [secret](/docs/integrations/secrets), injects it into requests at the network layer, and never exposes it to the model.

This also covers your own internal APIs and anything else that authenticates with a header.

When a service offers keys or tokens but not OAuth or MCP, paste the key into a secure card. Skydive stores it as a [secret](/docs/integrations/secrets), injects it into requests at the network layer, and never exposes it to the model.

**1. Generate a key from the service**

Most services have an API keys or developer settings section. Create a key there and copy it.

**2. Add the connection**

From the agent's connections, choose the service and paste the key when prompted.

**3. You're connected**

The key is stored securely. The agent uses it on every request without seeing the raw value.

> [!TIP]
>
> Keep the scope tight when the service lets you. An agent that only reads data doesn't need write access. If you need to rotate the key later, paste the new one in the same place and the old one stops working.

## Calling an API directly

With the key stored, the agent calls the service the normal way and the proxy attaches the credential in flight. It does not need a wrapper or an SDK, and it never handles the raw value.

## Managing a connection

Once a service is connected, manage it from the integration's card in your workspace. The card shows which agents use the connection and gives you two options to pull access back:

* **Remove from one agent.** The connection stays live for your other agents.
* **Disconnect entirely.** This revokes Skydive's access at the provider and removes it from every agent using it.

> **[Catalog](/docs/integrations/catalog)**
>
> Every service available to connect.

> **[Permissions](/docs/integrations/permissions)**
>
> How credentials stay out of the model's reach.


# Apollo
Source: https://getskydive.com/docs/integrations/apollo

Let your agent find and enrich leads in Apollo.io

Apollo.io is a sales intelligence platform for finding and reaching buyers. Connect it and your agent can search people and companies, enrich contacts, and work with your Apollo records.

## Two ways to connect

Apollo offers both an MCP connection and OAuth. MCP is the preferred path for most workflows.

* **Apollo MCP (recommended).** Direct access to Apollo's sales intelligence tools through the hosted MCP server. See [Connect with MCP](#connect-with-mcp).
* **Apollo OAuth.** The original sign-in integration. See [Connect with OAuth](#connect-with-oauth).

## What the agent can do

> **Search people**
>
> Find prospects by title, company, location, and more.

> **Enrich contacts**
>
> Pull emails, LinkedIn, and other firmographic detail.

> **Work accounts**
>
> Look up organizations and keep account records current.

> **Manage lists**
>
> Search contacts and accounts you already track in Apollo.

## Example agent tasks

```text theme={null}
Find 20 VPs of Sales at B2B SaaS companies in the US with 50 to 500 employees and export name, company, and work email.
```

```text theme={null}
Enrich this spreadsheet of leads with title, company size, and LinkedIn URL.
```

```text theme={null}
Pull the contacts on the Acme account and draft a short intro note for each.
```

## Connect with MCP

**1. Add the Apollo connection**

From the agent's connections, choose Apollo. The platform connects you through Apollo's MCP server.

**2. Sign in to Apollo**

A connect card appears. Sign in to your Apollo account and approve access.

**3. The agent gets to work**

It can now search, enrich, and work with your Apollo data through the MCP server.

## Connect with OAuth

## What you share

During the connect flow you authorize what the agent can reach. It works only within what you grant. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

**1. Authorize Apollo for the agent**

From the agent's connections, choose Apollo and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[Apollo API docs](https://docs.apollo.io/)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# GraphOS MCP Tools
Source: https://getskydive.com/docs/integrations/apollographql

GraphQL API platform and schema management

## Example agent tasks

```text theme={null}
Connect GraphOS MCP Tools and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from GraphOS MCP Tools.
```

```text theme={null}
Use the agent to interact with GraphOS MCP Tools directly from your conversation.
```

## Connecting

Connect GraphOS Tools over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add GraphOS MCP Tools.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach GraphOS MCP Tools through the MCP server.

## What you share

Connecting grants the agent access to GraphOS MCP Tools on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Apple Music
Source: https://getskydive.com/docs/integrations/apple

Apple's music streaming service

## What the agent can do

> **Sign-in**
>
> Do not invoke in freeform chat.

> **Search**
>
> Capabilities: Search the Apple Music catalog for artists, songs, albums, and playlists using natural-language queries.

> **Get-track-details-batch**
>
> Description: Batch-retrieves full song resource objects for a provided list of song titles and (optionally) artists.

> **Create-playlist-from-ids**
>
> Description: Creates a new user playlist from a list of song IDs.

## Example agent tasks

```text theme={null}
Search Apple Music for relevant results and summarize what you find.
```

```text theme={null}
Pull the latest data from Apple Music and give me a summary.
```

```text theme={null}
Create a new entry in Apple Music based on my notes.
```

## Connecting

Connect Apple Music over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Apple Music.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Apple Music through the MCP server.

## What you share

Connecting grants the agent access to Apple Music on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# ApplyBoard MCP
Source: https://getskydive.com/docs/integrations/applyboard

International student university application platform

## Example agent tasks

```text theme={null}
Connect ApplyBoard MCP and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from ApplyBoard MCP.
```

```text theme={null}
Use the agent to interact with ApplyBoard MCP directly from your conversation.
```

## Connecting

Connect ApplyBoard over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add ApplyBoard.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach ApplyBoard through the MCP server.

## What you share

Connecting grants the agent access to ApplyBoard MCP on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Archive
Source: https://getskydive.com/docs/integrations/archive

Influencer and UGC content archive for Instagram, TikTok, and YouTube

Archive captures the social content your brand is part of: every Instagram, TikTok, and YouTube post that mentions or is paid for by your team lands in one searchable library of creators, engagement history, and saved views. Connect it and your agents can search that library, answer questions about creators and performance, and manage the views and collections your team organizes content into.

## What your agents can do

> **Search captured content**
>
> Find Instagram, TikTok, and YouTube posts in the archive by creator, campaign, date, or saved view.

> **Read creators and engagement**
>
> Pull creator contact details and engagement history across everything the workspace has captured.

> **Manage views and collections**
>
> Create and update the saved views and collections your team uses to group content.

> **Work the GraphQL API directly**
>
> Use an API token to query the full GraphQL endpoint for anything the MCP tools do not cover.

## Example agent tasks

```text theme={null}
Find every TikTok post we captured from this creator last quarter and summarize engagement.
```

```text theme={null}
Build a new collection from the top 20 UGC posts by view count this month.
```

```text theme={null}
List the creators in our Q4 campaign views and export their contact details.
```

## What you share

Connecting over MCP gives your agents read access to the archive: captured posts, creators, engagement history, and saved views. Read-only covers 32 tools; if you grant the write scope during consent, your agents can also manage collections, views, webhooks, and imports (31 more tools). Deletes only remove the saved view or collection itself, never the underlying content or creators. Every tool call spends the workspace's shared API credit budget, same as the GraphQL API, so a long agent task competes with your own integrations for that budget. See [Permissioning](/docs/integrations/permissions).

## Connecting

**MCP server**

Connect Archive over MCP to search archived social content, read creators and engagement history, and manage saved views and collections.

**1. Add the connection**

From your agent's connections, choose Archive and sign in to your Archive account.

**2. Choose read-only or write access**

Archive asks whether to also grant write access. Leave it off for read-only; turn it on only if your agents should manage views, collections, and imports.

**3. Your agents get to work**

Start with getWorkspaces to verify the session and get the workspace IDs the other tools need.

**API key**

Connect Archive through its GraphQL API with a workspace API token.

**1. Create an API token**

In the Archive app, open the Integration tab and create an API token.

**2. Add the Archive connection**

From your agent's connections, choose Archive and paste your token when prompted.

**3. Your agents get to work**

Your agents can now query the GraphQL API at app.archive.com/api/v2, from searching posts to reading creator details.

## Learn more

> **[Archive](https://archive.com)**
>
> Product overview and pricing.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Arcjet
Source: https://getskydive.com/docs/integrations/arcjet

Bot detection, rate limiting and app security

## Example agent tasks

```text theme={null}
Connect Arcjet and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Arcjet.
```

```text theme={null}
Use the agent to interact with Arcjet directly from your conversation.
```

## Connecting

Connect Arcjet over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Arcjet.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Arcjet through the MCP server.

## What you share

Connecting grants the agent access to Arcjet on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Asana
Source: https://getskydive.com/docs/integrations/asana

Let your agents manage your Asana tasks and projects

Asana tracks team work in projects and tasks. Connect it and your agents can read your projects, create tasks, and update them.

## What your agents can do

> **Read tasks**
>
> See what's in a project and what's due.

> **Create tasks**
>
> Add tasks to a project.

> **Update tasks**
>
> Change status, assignee, and due date.

> **See projects**
>
> Find the right project to work in.

## Example agent tasks

```text theme={null}
When a customer request comes in, file it as a task in the right project with a due date.
```

```text theme={null}
Every Monday, tell me what's due this week and what's already slipping.
```

```text theme={null}
When I mark a task done, draft the status update to the stakeholder it affects.
```

## What you share

During the connect flow you authorize what the agent can reach. It works only within what you grant. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Authorize Asana for the agent**

From the agent's connections, choose Asana and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[Asana API docs](https://developers.asana.com/docs)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Ashby
Source: https://getskydive.com/docs/integrations/ashby

All-in-one ATS with built-in sourcing and analytics

Ashby is an all-in-one applicant tracking system with sourcing and analytics built in. Connect it and your agent can search candidates and pipelines, pull interview schedules and job metrics, and take action inside your ATS without switching tabs.

## What the agent can do

> **Search candidates and jobs**
>
> Find candidates by name, stage, or job, and pull details on applications and interview plans.

> **Track pipeline metrics**
>
> Pull funnel data, conversion rates, and pipeline summaries across jobs and time ranges.

> **Pull interview schedules**
>
> Check upcoming interviews, interviewer assignments, and availability for any requisition.

> **Write to the ATS**
>
> Add candidate notes, create new candidates, attach them to jobs, and move applications between stages.

## Example agent tasks

```text theme={null}
Show me all candidates in the final round for the senior engineer role.
```

```text theme={null}
Add a note to Jordan Kim's application: strong culture fit, advance to offer.
```

```text theme={null}
What's the pipeline conversion rate for our product manager openings this quarter?
```

## What you share

Connecting gives the agent access to your Ashby workspace, including candidate records, job details, interview plans, and pipeline data. The agent can read and write within the scope your account allows. See [Permissions](/docs/integrations/permissions).

## Connecting

**1. Add the Ashby connection**

From the agent's connections, choose Ashby. The platform connects you through Ashby's MCP server.

**2. Sign in to Ashby**

A connect card appears. Sign in to your Ashby account and approve access.

**3. The agent gets to work**

It can now search candidates, pull pipeline metrics, and update records in your ATS.

## Learn more

> **[Ashby](https://docs.ashbyhq.com)**
>
> Ashby documentation and help center.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Atlassian
Source: https://getskydive.com/docs/integrations/atlassian

Let your agent work across Jira, Confluence, and more

Atlassian ties together Jira, Confluence, and more under one sign-in. Connect it and your agent can manage issues, read and write docs, and search across your Atlassian products.

## What the agent can do

> **Jira issues**
>
> Create, read, edit, and transition Jira issues. Search with JQL.

> **Confluence pages**
>
> Read and create Confluence pages, blog posts, and live docs. Search with CQL.

> **Cross-product search**
>
> Find content across Jira and Confluence in one query.

> **Loom videos**
>
> Pull video metadata, transcripts, and AI briefs from linked Loom recordings.

## Example agent tasks

```text theme={null}
When a customer files a ticket, open the Jira issue and link the relevant Confluence runbook.
```

```text theme={null}
Every Monday, give me one rollup of what's at risk across Jira and what docs went stale.
```

```text theme={null}
Trace this bug from the issue to the doc that should have prevented it.
```

## What you share

During the connect flow you authorize what the agent can reach. It works only within what you grant. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Add the Atlassian connection**

From the agent's connections, choose Atlassian. The platform connects you through Atlassian's official MCP server.

**2. Sign in to Atlassian**

A connect card appears. Sign in to your Atlassian account and approve access to your Jira and Confluence products.

**3. The agent gets to work**

It can now manage issues, read and write Confluence content, and search across your Atlassian workspace.

## Learn more

> **[Atlassian developer docs](https://developer.atlassian.com/cloud/)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Attio
Source: https://getskydive.com/docs/integrations/attio

Let your agent work with your Attio CRM

Attio is a flexible, modern CRM. Connect it and your agent can read your records and create or update them.

## What the agent can do

> **Read records**
>
> Pull people, companies, and deals.

> **Create records**
>
> Add new records.

> **Update records**
>
> Change attributes on a record.

> **Search**
>
> Find the right record.

## Example agent tasks

```text theme={null}
When I add a prospect, enrich the record and draft a first outreach note.
```

```text theme={null}
Each Monday, tell me which deals have gone quiet and need a touch.
```

```text theme={null}
Log this meeting to the right record and set the next follow-up.
```

## What you share

During the connect flow you authorize what the agent can reach. It works only within what you grant. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Authorize Attio for the agent**

From the agent's connections, choose Attio and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[Attio API docs](https://developers.attio.com/reference)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# AuthZed MCP Server
Source: https://getskydive.com/docs/integrations/authzed

Fine-grained authorization and permissions engine

## Example agent tasks

```text theme={null}
Connect AuthZed MCP Server and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from AuthZed MCP Server.
```

```text theme={null}
Use the agent to interact with AuthZed MCP Server directly from your conversation.
```

## Connecting

Connect AuthZed over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add AuthZed.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach AuthZed through the MCP server.

## What you share

Connecting grants the agent access to AuthZed MCP Server on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Autodesk
Source: https://getskydive.com/docs/integrations/autodesk

3D design, engineering and construction software

## Example agent tasks

```text theme={null}
Connect Autodesk and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Autodesk.
```

```text theme={null}
Use the agent to interact with Autodesk directly from your conversation.
```

## Connecting

Connect Autodesk over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Autodesk.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Autodesk through the MCP server.

## What you share

Connecting grants the agent access to Autodesk on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Avo
Source: https://getskydive.com/docs/integrations/avo

Analytics tracking plan and data governance

## Example agent tasks

```text theme={null}
Connect Avo and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Avo.
```

```text theme={null}
Use the agent to interact with Avo directly from your conversation.
```

## Connecting

Connect Avo over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Avo.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Avo through the MCP server.

## What you share

Connecting grants the agent access to Avo on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Avoma
Source: https://getskydive.com/docs/integrations/avoma

Let your agent work with your Avoma meeting recordings and insights

Avoma records and transcribes meetings, surfaces action items, and helps sales and customer teams track conversation quality. Connect it and your agent can pull meeting summaries, track follow-ups, and surface patterns across your recorded conversations.

## What the agent can do

> **Pull meeting summaries**
>
> Retrieve AI-generated summaries and action items from recorded meetings.

> **Search transcripts**
>
> Find specific moments or topics across your call library.

> **Track follow-ups**
>
> Surface action items and next steps from recent meetings.

> **Review call insights**
>
> Pull talking time ratios, topic coverage, and engagement metrics.

## Example agent tasks

```text theme={null}
Summarize the last five calls with the Acme account and list open action items.
```

```text theme={null}
Which meetings from this week had action items that haven't been completed yet?
```

```text theme={null}
Pull the transcript from Monday's discovery call and highlight the customer's stated requirements.
```

## What you share

Connecting gives the agent access to your Avoma workspace, including meeting recordings, transcripts, and insights within your account permissions. Requires an Avoma Organization or Enterprise plan, or the API Access add-on. See [Permissions](/docs/integrations/permissions).

## Connecting

**1. Add the Avoma connection**

From the agent's connections, choose Avoma. The platform connects you through Avoma's MCP server.

**2. Sign in to Avoma**

A connect card appears. Sign in to your Avoma account and approve access.

**3. The agent gets to work**

It can now pull meeting data, transcripts, and insights from your Avoma workspace.

## Learn more

> **[Avoma help center](https://help.avoma.com)**
>
> Product guides and the Avoma API reference.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Axiom
Source: https://getskydive.com/docs/integrations/axiom

Log management and observability platform

## Example agent tasks

```text theme={null}
Connect Axiom and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Axiom.
```

```text theme={null}
Use the agent to interact with Axiom directly from your conversation.
```

## Connecting

Connect Axiom over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Axiom.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Axiom through the MCP server.

## What you share

Connecting grants the agent access to Axiom on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# B12 MCP server
Source: https://getskydive.com/docs/integrations/b12

AI-built websites for small businesses

## What the agent can do

> **Generate Website**
>
> Generate a website with AI.

## Example agent tasks

```text theme={null}
Ask the agent to help you with B12 MCP server tasks using natural language.
```

## Connecting

Connect B12 over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add B12.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach B12 through the MCP server.

## What you share

Connecting grants the agent access to B12 MCP server on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Backstage MCP server
Source: https://getskydive.com/docs/integrations/backstage

Developer portal for service catalogs and docs

## Example agent tasks

```text theme={null}
Connect Backstage MCP server and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Backstage MCP server.
```

```text theme={null}
Use the agent to interact with Backstage MCP server directly from your conversation.
```

## Connecting

Connect Backstage over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Backstage.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Backstage through the MCP server.

## What you share

Connecting grants the agent access to Backstage MCP server on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Ballpark
Source: https://getskydive.com/docs/integrations/ballparkhq

Async video interviews for hiring teams

## Example agent tasks

```text theme={null}
Connect Ballpark and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Ballpark.
```

```text theme={null}
Use the agent to interact with Ballpark directly from your conversation.
```

## Connecting

Connect Ballpark over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Ballpark.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Ballpark through the MCP server.

## What you share

Connecting grants the agent access to Ballpark on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Basecamp
Source: https://getskydive.com/docs/integrations/basecamp

Let your agent work with your Basecamp projects

Basecamp keeps projects, to-dos, and messages in one place. Connect it and your agent can read your projects and post updates.

## What the agent can do

> **Read to-dos**
>
> See what's on a project's to-do lists.

> **Add to-dos**
>
> Create new to-do items.

> **Read messages**
>
> Catch up on a project's message board.

> **See projects**
>
> Find the right project.

## Example agent tasks

```text theme={null}
Turn the latest client message into to-dos assigned to the right people.
```

```text theme={null}
Every Friday, summarize each project's progress for the client update.
```

```text theme={null}
Flag any to-do that's overdue and draft the nudge to its owner.
```

## What you share

During the connect flow you authorize what the agent can reach. It works only within what you grant. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Authorize Basecamp for the agent**

From the agent's connections, choose Basecamp and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[Basecamp API docs](https://github.com/basecamp/bc3-api)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Bitbucket
Source: https://getskydive.com/docs/integrations/bitbucket

Let your agent work with your Bitbucket repos

Bitbucket hosts your Git repositories and pull requests. Connect it and your agent can read and write across the repos you grant.

## What the agent can do

> **Read and write code**
>
> Work with files in a repository.

> **Pull requests**
>
> Open and review pull requests.

> **Issues**
>
> Create and update issues.

> **Branches**
>
> Work with branches and commits.

## Example agent tasks

```text theme={null}
Open a pull request from this branch and write the description from the commits.
```

```text theme={null}
Every Monday, list the PRs that have been open too long and ping their authors.
```

```text theme={null}
Tell me what changed in the last release and whether anything looks risky.
```

## What you share

During the connect flow you authorize what the agent can reach. It works only within what you grant. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Authorize Bitbucket for the agent**

From the agent's connections, choose Bitbucket and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[Bitbucket Cloud REST docs](https://developer.atlassian.com/cloud/bitbucket/rest/intro/)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Bitly
Source: https://getskydive.com/docs/integrations/bitly

Let your agent shorten links and read click stats

Bitly shortens links and tracks how they perform. Connect it and your agent can create short links, generate QR codes, and pull click analytics.

## Two ways to connect

Bitly now runs a hosted MCP server. Connect it and the agent can manage your links and pull analytics without needing a separate API setup.

* **Bitly MCP (recommended).** Connect Bitly's MCP server once and the agent can create short links and QR codes, pull click analytics, manage your link library. See [Connect with MCP](#connect-with-mcp).
* **Bitly OAuth.** The original sign-in integration, scoped to what you grant. See [Connect with OAuth](#connect-with-oauth).

## What the agent can do

> **Create short links**
>
> Turn long URLs into short, shareable links.

> **Generate QR codes**
>
> Create QR codes tied to your short links.

> **Pull click analytics**
>
> See clicks by time, country, city, device, and referrer.

> **Manage links**
>
> Update destinations, titles, tags, and archive old links.

## Example agent tasks

```text theme={null}
Create a short link for our new product page and track how it performs over the first week.
```

```text theme={null}
Generate a QR code for this event and send me the image.
```

```text theme={null}
Which of our campaign links got the most clicks this month?
```

## Connect with MCP

**1. Add the Bitly connection**

From the agent's connections, choose Bitly. The platform connects you through Bitly's MCP server.

**2. Sign in to Bitly**

A connect card appears. Sign in to your Bitly account and approve.

**3. The agent gets to work**

It can now create short links and QR codes, pull click analytics, manage your link library through the MCP server.

> [!NOTE]
>
> MCP is the first-class connection when both options are available. See [Tools & MCP](/docs/capabilities/tools-and-mcp).

## Connect with OAuth

## What you share

During the connect flow you authorize what the agent can reach. It works only within what you grant. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

**1. Authorize Bitly for the agent**

From the agent's connections, choose Bitly and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[Bitly developer docs](https://dev.bitly.com)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Blockscout
Source: https://getskydive.com/docs/integrations/blockscout

Open-source blockchain explorer and analytics

## Example agent tasks

```text theme={null}
Connect Blockscout and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Blockscout.
```

```text theme={null}
Use the agent to interact with Blockscout directly from your conversation.
```

## Connecting

Connect Blockscout over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Blockscout.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Blockscout through the MCP server.

## What you share

Connecting grants the agent access to Blockscout on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# viaNexus vAST
Source: https://getskydive.com/docs/integrations/blueskyapi

Agent access layer for institutional financial data

## Example agent tasks

```text theme={null}
Connect viaNexus vAST and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from viaNexus vAST.
```

```text theme={null}
Use the agent to interact with viaNexus vAST directly from your conversation.
```

## Connecting

Connect viaNexus vAST over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add viaNexus vAST.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach viaNexus vAST through the MCP server.

## What you share

Connecting grants the agent access to viaNexus vAST on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# BlueSuite
Source: https://getskydive.com/docs/integrations/bluesuite

Job management software for home service pros

## Example agent tasks

```text theme={null}
Connect BlueSuite and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from BlueSuite.
```

```text theme={null}
Use the agent to interact with BlueSuite directly from your conversation.
```

## Connecting

Connect BlueSuite over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add BlueSuite.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach BlueSuite through the MCP server.

## What you share

Connecting grants the agent access to BlueSuite on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Box
Source: https://getskydive.com/docs/integrations/box

Let your agents work with your files in Box

Box stores your files and folders in the cloud. Connect it and your agents can find, read, and upload files.

## Two ways to connect

Box now runs a hosted MCP server. For most people, that is the easier path: one connection, then file and folder tools across the content the signed-in user can already reach.

* **Box MCP server (recommended).** Connect Box's hosted MCP server once and your agents can search, read, and manage files and folders. Your Box admin must enable the Box MCP server in the Admin Console first. See [Connect with MCP](#connect-with-mcp).
* **Box OAuth.** The original sign-in integration, scoped to what you grant. See [Connect with OAuth](#connect-with-oauth).

## What your agents can do

> **Find files**
>
> Search your Box account.

> **Read files**
>
> Pull content from a document.

> **Upload files**
>
> Save new files to a folder.

> **Browse folders**
>
> Move through your folder structure.

## Example agent tasks

```text theme={null}
When a signed contract lands, file it in the right client folder and log the renewal date.
```

```text theme={null}
Every week, gather the new files across my projects into a single digest.
```

```text theme={null}
Find every document mentioning the old pricing and list what needs updating.
```

## Connect with MCP

Connect to Box's hosted MCP server to search, read, and manage files and folders in your Box account. Box's own hosted MCP gives agents direct file and folder tools.

> [!NOTE]
>
> Requires a Box account whose admin has enabled the Box MCP server in the Admin Console (Integrations, then MCP). Tools only reach content the signed-in user can already access. Write tools change real files, so confirm before destructive changes.

**1. Add the Box connection**

From the agent's connections, choose Box. The platform connects you through Box's MCP server.

**2. Sign in to Box**

A connect card appears. Sign in to your Box account and approve.

**3. Your agents get to work**

They can now search, read, and manage files and folders through the MCP server, within what the signed-in user can already access.

> [!NOTE]
>
> MCP is the first-class connection when both options are available. See [Tools & MCP](/docs/capabilities/tools-and-mcp).

## Connect with OAuth

## What you share

During the connect flow you authorize what the agent can reach. It works only within what you grant. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

**1. Authorize Box for the agent**

From the agent's connections, choose Box and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[Box MCP server](https://developer.box.com/guides/box-mcp/remote/)**
>
> How to enable the hosted MCP endpoint and which tools it exposes.

> **[Box API docs](https://developer.box.com/reference/)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Braintrust
Source: https://getskydive.com/docs/integrations/braintrust

Evaluation and observability platform for LLM apps

## Example agent tasks

```text theme={null}
Connect Braintrust and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Braintrust.
```

```text theme={null}
Use the agent to interact with Braintrust directly from your conversation.
```

## Connecting

**MCP server**

Connect Braintrust over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Braintrust.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Braintrust through the MCP server.

**API key**

Connect Braintrust with an API key.

**1. Add the connection**

Generate an API key from your Braintrust account settings.

**2. Authorize**

Copy the key. It is only shown once.

**3. The agent gets to work**

Paste it when prompted in the agent's connection card for Braintrust.

## What you share

Connecting grants the agent access to Braintrust on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Brevo
Source: https://getskydive.com/docs/integrations/brevo

Email marketing and CRM platform

## Example agent tasks

```text theme={null}
Connect Brevo and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Brevo.
```

```text theme={null}
Use the agent to interact with Brevo directly from your conversation.
```

## Connecting

Connect Brevo over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Brevo.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Brevo through the MCP server.

## What you share

Connecting grants the agent access to Brevo on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Brex
Source: https://getskydive.com/docs/integrations/brex

Corporate cards and spend management

## Example agent tasks

```text theme={null}
Connect Brex and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Brex.
```

```text theme={null}
Use the agent to interact with Brex directly from your conversation.
```

## Connecting

Connect Brex over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Brex.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Brex through the MCP server.

## What you share

Connecting grants the agent access to Brex on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Browser
Source: https://getskydive.com/docs/integrations/browser

For services with no OAuth, MCP, or API key

Some services have no OAuth, no MCP server, and no API. The agent can still use them, by signing in through the [shared browser](/docs/capabilities/browser) and holding the session.

This is the fallback that makes "anything you can do, your agent can do" true rather than aspirational.

## How it works

**1. The agent opens the site**

It navigates to the sign-in page in its own browser and stops there.

**2. You sign in yourself**

The agent sends you a link to watch and control that browser. You enter the password, the code, or whatever the login needs. The credential never passes through the agent or the conversation.

**3. It carries on**

The session is held in the browser, and the agent continues the task from where it stopped.

For services that support neither MCP, OAuth, nor API keys, the agent can use the [shared browser](/docs/capabilities/browser). You log in yourself in the browser. The session is captured so the agent can keep working in that site.

## When to use it

Reach for this when there is genuinely no programmatic option, or when a login must be yours: an SSO flow, a CAPTCHA, a two-factor code. Prefer [OAuth](/docs/integrations/oauth), [MCP](/docs/integrations/mcp), or an [API key](/docs/integrations/apis) whenever the service offers one, since a held session is more fragile than a token.

> **[Browser use](/docs/capabilities/browser)**
>
> What the agent's browser can do, and how the shared view works.

> **[Catalog](/docs/integrations/catalog)**
>
> Check whether the service has a cleaner method first.


# Buffer
Source: https://getskydive.com/docs/integrations/buffer

Let your agent schedule and manage posts in Buffer

Buffer is a social media scheduler. Connect it and your agent can create posts, manage ideas, and work with your connected channels.

## Two ways to connect

Buffer supports sign-in with OAuth and a personal API key. OAuth is the default: one click, then the agent is ready. The API key option is there if you prefer to use a key you already manage.

* **Buffer OAuth (recommended).** Sign in once and the agent connects through the secure proxy. See [Connect with OAuth](#connect-with-oauth).
* **Buffer API key.** Paste a personal API key from Buffer's settings. See [Connect with an API key](#connect-with-an-api-key).

## What the agent can do

> **Schedule posts**
>
> Create and queue posts to your channels.

> **Manage ideas**
>
> Save content ideas for later.

> **See channels**
>
> List the social profiles you have connected.

> **Read posts**
>
> Pull scheduled, draft, and sent posts.

## Example agent tasks

```text theme={null}
Queue this launch update across my Buffer channels for Tuesday at 10am.
```

```text theme={null}
Every Monday, draft three post ideas from last week's wins and leave them in Buffer for me to review.
```

```text theme={null}
List what's scheduled for the rest of the week and flag any gaps.
```

## Connect with OAuth

## What you share

During the connect flow you authorize what the agent can reach. It works only within what you grant. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

**1. Authorize Buffer for the agent**

From the agent's connections, choose Buffer and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Connect with an API key

**1. Get your Buffer API key**

In Buffer, go to [Settings → API](https://publish.buffer.com/settings/api) and create a new personal API key.

**2. Add the key to the agent**

From the agent's connections, choose Buffer and paste the key when prompted.

**3. The agent gets to work**

It uses the key to reach the Buffer API on your behalf.

## Learn more

> **[Buffer API docs](https://developers.buffer.com/)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Builder Content
Source: https://getskydive.com/docs/integrations/builder

Visual CMS for building web pages

## Example agent tasks

```text theme={null}
Connect Builder Content and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Builder Content.
```

```text theme={null}
Use the agent to interact with Builder Content directly from your conversation.
```

## Connecting

Connect Builder Content with an API key.

**1. Add the connection**

Generate an API key from your Builder Content account settings.

**2. Authorize**

Copy the key. It is only shown once.

**3. The agent gets to work**

Paste it when prompted in the agent's connection card for Builder Content.

## What you share

Connecting grants the agent access to Builder Content on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Fern Docs MCP server
Source: https://getskydive.com/docs/integrations/buildwithfern

API documentation generation platform

## Example agent tasks

```text theme={null}
Connect Fern Docs MCP server and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Fern Docs MCP server.
```

```text theme={null}
Use the agent to interact with Fern Docs MCP server directly from your conversation.
```

## Connecting

Connect Fern Docs over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Fern Docs.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Fern Docs through the MCP server.

## What you share

Connecting grants the agent access to Fern Docs MCP server on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Bump.sh APIs hub MCP server
Source: https://getskydive.com/docs/integrations/bump

Hosted API documentation and reference hub

## Example agent tasks

```text theme={null}
Connect Bump.sh APIs hub MCP server and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Bump.sh APIs hub MCP server.
```

```text theme={null}
Use the agent to interact with Bump.sh APIs hub MCP server directly from your conversation.
```

## Connecting

Connect Bump.sh Workspace docs over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Bump.sh APIs hub.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Bump.sh APIs hub through the MCP server.

## What you share

Connecting grants the agent access to Bump.sh APIs hub MCP server on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Cal.com
Source: https://getskydive.com/docs/integrations/cal-com

Let your agent manage your Cal.com scheduling

Cal.com is open-source scheduling. Connect it and your agent can read your bookings and event types and help manage your calendar of meetings.

## What the agent can do

> **Read bookings**
>
> See scheduled meetings and their details.

> **Event types**
>
> Read your meeting types.

> **See invitees**
>
> Pull who booked and when.

> **Find events**
>
> Search your scheduled events.

## Example agent tasks

```text theme={null}
Each morning, brief me on today's Cal.com meetings and who I'm meeting.
```

```text theme={null}
When someone books a demo, pull their details and prep me beforehand.
```

```text theme={null}
Every Friday, tell me how my week booked out.
```

## What you share

During the connect flow you authorize what the agent can reach. It works only within what you grant. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Authorize Cal.com for the agent**

From the agent's connections, choose Cal.com and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[Cal.com API docs](https://cal.com/docs/api-reference)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Calendly
Source: https://getskydive.com/docs/integrations/calendly

Let your agent read your Calendly scheduling

Calendly handles scheduling. Connect it and your agent can check your availability, create bookings, and manage your event types.

## Two ways to connect

Calendly now runs a hosted MCP server. Connect it and the agent can work with your schedule directly.

* **Calendly MCP (recommended).** Connect Calendly's MCP server once and the agent can read events and availability, create and cancel bookings, manage event types. See [Connect with MCP](#connect-with-mcp).
* **Calendly OAuth.** The original sign-in integration, scoped to what you grant. See [Connect with OAuth](#connect-with-oauth).

## What the agent can do

> **Check availability**
>
> See open time slots and busy periods.

> **Create bookings**
>
> Book meetings on your behalf.

> **Manage event types**
>
> Read and update your event type settings.

> **Cancel meetings**
>
> Cancel scheduled events.

## Example agent tasks

```text theme={null}
Check my availability this week and share a booking link with the team.
```

```text theme={null}
Cancel my 3pm tomorrow and find the next open slot for a follow-up.
```

```text theme={null}
List all the meetings I have booked for next week.
```

## Connect with MCP

**1. Add the Calendly connection**

From the agent's connections, choose Calendly. The platform connects you through Calendly's MCP server.

**2. Sign in to Calendly**

A connect card appears. Sign in to your Calendly account and approve.

**3. The agent gets to work**

It can now read events and availability, create and cancel bookings, manage event types through the MCP server.

> [!NOTE]
>
> MCP is the first-class connection when both options are available. See [Tools & MCP](/docs/capabilities/tools-and-mcp).

## Connect with OAuth

## What you share

During the connect flow you authorize what the agent can reach. It works only within what you grant. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

**1. Authorize Calendly for the agent**

From the agent's connections, choose Calendly and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[Calendly API docs](https://developer.calendly.com/api-docs)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Canva
Source: https://getskydive.com/docs/integrations/canva

Online graphic design and visual content tool

## Example agent tasks

```text theme={null}
Connect Canva and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Canva.
```

```text theme={null}
Use the agent to interact with Canva directly from your conversation.
```

## Connecting

Connect Canva over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Canva.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Canva through the MCP server.

## What you share

Connecting grants the agent access to Canva on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Capacities
Source: https://getskydive.com/docs/integrations/capacities

Let your agent work with your Capacities knowledge base

Capacities is a personal knowledge management tool organized around object types rather than folders. Connect it and your agent can create notes, search your space, and pull structured content from your knowledge base.

## What the agent can do

> **Create notes and objects**
>
> Add new items to your Capacities space with the right type and properties.

> **Search your space**
>
> Find notes, people, books, and other objects across your knowledge base.

> **Read structured content**
>
> Pull properties and relationships from your Capacities objects.

> **Update existing content**
>
> Edit notes and object properties inside your space.

## Example agent tasks

```text theme={null}
Create a new person entry for the contact I just met, with their company and notes from our conversation.
```

```text theme={null}
Find all book notes I've taken in the past month and list their key takeaways.
```

```text theme={null}
Add today's meeting notes to my Capacities space under the right project.
```

## What you share

Connecting gives the agent access to your Capacities space, including notes, objects, and their properties. Requires a Capacities Pro plan. See [Permissions](/docs/integrations/permissions).

## Connecting

**1. Add the Capacities connection**

From the agent's connections, choose Capacities. The platform connects you through Capacities' hosted MCP server.

**2. Sign in to Capacities**

A connect card appears. Sign in to your Capacities account and approve access. A Pro plan is required.

**3. The agent gets to work**

It can now read and write your Capacities knowledge base.

## Learn more

> **[Capacities documentation](https://docs.capacities.io)**
>
> Guides and the Capacities object model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Carta
Source: https://getskydive.com/docs/integrations/carta

Equity, cap table, and fund administration, including the Carta CRM

Carta covers equity, cap table, and fund administration, and its CRM manages contacts, companies, deals, and fundraising opportunities for firms. Connect the Carta CRM MCP server and your agents can search records, work deals through pipelines, write notes, and run reports, with 91 read and 67 write tools scoped to what your organization has enabled.

## What your agents can do

> **Search and enrich records**
>
> Search contacts, companies, deals, investors, and fundraising opportunities by name, domain, or text query. enrich\_person and enrich\_contact pull in outside data.

> **Work deals and pipelines**
>
> Move deals through stages, read stage history, and aggregate deals by field for pipeline reporting.

> **Keep records current**
>
> Create and update contacts, companies, notes, and tasks. Write tools touch real records, so agents confirm before deleting or merging.

> **Run reports and exports**
>
> List and read saved reports, schedule them, and generate CSV exports or PDFs from entity records.

## Example agent tasks

```text theme={null}
Search Carta CRM for this contact, pull their recent interactions and notes, and draft a follow-up task.
```

```text theme={null}
Find companies in our pipeline that have not had an interaction in 30 days and list them by stage.
```

```text theme={null}
Create a contact from this email signature, add them to the Founders list, and link a note explaining the intro.
```

## Connecting

**MCP server**

Connect to the Carta CRM through its official MCP server with your Carta account. Tools only return data for modules your organization has enabled and records you can access.

**1. Add the connection**

From your agent's connections, choose Carta and sign in with the Carta account for your firm.

**2. The agent gets to work**

Start with search\_contacts or search\_deals, and read get\_current\_user plus get\_tenant\_custom\_instructions once per session for your organization's agent guidance.

Write tools create and update real records in the Carta CRM. Deleting or merging contacts is a destructive change, so agents confirm with you first.

## What you share

Connecting grants your agents access to the Carta CRM on your behalf, scoped to your Carta account. They can read and write contacts, companies, deals, fundraising opportunities, investors, notes, and tasks. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

## Learn more

> **[Carta CRM MCP documentation](https://docs.carta.com/crm/docs/carta-crm-mcp)**
>
> The full tool list for the Carta CRM MCP server, read and write.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Catalog
Source: https://getskydive.com/docs/integrations/catalog

The services and tools your agent can connect to

Skydive agents connect to outside services through OAuth, MCP, API keys, and Portal. Stored credentials stay behind a secure [proxy](/docs/integrations/permissions), while [Portal](/docs/integrations/portal) connects an agent to tools running on your computer.

This page lists the services in the catalog. It grows over time, and for anything not here yet you can still grant access with an [API key or secret](/docs/integrations/secrets).

## Every connection widens what your agent owns

One integration is useful. Several together is what lets an agent own a real slice of the work. Each connection you add is another part of your day it can act on, and it can move across all of them in a single task.

* Close a deal in your CRM, raise the invoice in your accounting tool, and start the onboarding checklist, in one go.
* Watch your support inbox, file the bug in your issue tracker, and post a heads-up to the right channel.
* Pull last week's numbers from analytics, write the recap, and drop it in your docs every Monday morning.

An agent's reach comes down to what you connect and what each service's API can do. Connect more, and it does more. The agents that earn their keep are usually wired into several tools at once, so they handle the whole job instead of one step of it.

> [!TIP]
>
> Start with the one or two tools a task touches most. Add connections as the agent takes on more, and it starts joining the dots across them on its own.

> [!NOTE]
>
> New providers are added regularly. If a service you need isn't listed, an [API key or secret](/docs/integrations/secrets) covers it in the meantime.

## How connecting works

**1. Give your agent a task**

Hand it work that needs a service, like "check my calendar for tomorrow." If it needs access, the agent asks.

**2. Authorize with one click**

A connect card appears. Sign in to the provider and approve. The credential is stored securely.

**3. The agent gets to work**

Its requests to that service are authenticated automatically through the proxy. See [Permissioning](/docs/integrations/permissions).

> [!NOTE]
>
> Looking for Slack, email, or SMS/iMessage? Those are messaging [channels](/docs/channels/overview) your agent lives on, not OAuth integrations.

## Browse the catalog

Connect any of these from Skydive. More are on the way.

### Google Workspace

Gmail, Drive, Calendar, Docs, Sheets, Slides, Contacts, Tasks, and Forms. Google approved Skydive's OAuth client for these Workspace scopes. Connecting any of them locks that agent to [trusted models](/docs/agents/trusted-models). Overview: [Google Workspace](/docs/integrations/google).

> **[Gmail](/docs/integrations/gmail)**

> **[Drive](/docs/integrations/google-drive)**

> **[Calendar](/docs/integrations/google-calendar)**

> **[Docs](/docs/integrations/google-docs)**

> **[Sheets](/docs/integrations/google-sheets)**

> **[Slides](/docs/integrations/google-slides)**

> **[Contacts](/docs/integrations/google-contacts)**

> **[Tasks](/docs/integrations/google-tasks)**

> **[Forms](/docs/integrations/google-forms)**

### Other Google connectors

Live in product. Not part of the Workspace OAuth app-review scope set above.

> **[Ads](/docs/integrations/google-ads)**

> **[Analytics](/docs/integrations/google-analytics)**

> **[Meet](/docs/integrations/google-meet)**

> **[Play](/docs/integrations/google-play)**

> **[Search Console](/docs/integrations/google-search-console)**

### Productivity and docs

> **[Airtable](/docs/integrations/airtable)**

> **[Box](/docs/integrations/box)**

> **[Contentful](/docs/integrations/contentful)**

> **[Dropbox](/docs/integrations/dropbox)**

> **[Dropbox Sign](/docs/integrations/dropbox-sign)**

> **[Notion](/docs/integrations/notion)**

> **[SignNow](/docs/integrations/signnow)**

> **[Agility CMS](/docs/integrations/agilitycms)**

> **[B12 MCP server](/docs/integrations/b12)**

> **[Ballpark](/docs/integrations/ballparkhq)**

> **[BlueSuite](/docs/integrations/bluesuite)**

> **[Clarify](/docs/integrations/clarify)**

> **[Common Room](/docs/integrations/commonroom)**

> **[Cosmic Hosted MCP (agent scope)](/docs/integrations/cosmicjs)**

> **[Crossbeam](/docs/integrations/crossbeam)**

> **[Cube](/docs/integrations/cubesoftware)**

> **[Day AI](/docs/integrations/day)**

> **[Egnyte](/docs/integrations/egnyte)**

> **[Fathom](/docs/integrations/fathom)**

> **[Garchi CMS MCP OAuth Endpoint](/docs/integrations/garchi)**

> **[Gorgias](/docs/integrations/gorgias)**

> **[Guru MCP Server](/docs/integrations/getguru)**

> **[Hercules](/docs/integrations/hercules)**

> **[Jam](/docs/integrations/jam)**

> **[Jimdo](/docs/integrations/jimdo)**

> **[Kit](/docs/integrations/kit)**

> **[Klaviyo](/docs/integrations/klaviyo)**

> **[Krisp](/docs/integrations/krisp)**

> **[Lemlist](/docs/integrations/lemlist)**

> **[Linktree MCP Server](/docs/integrations/linktr)**

> **[Lumin](/docs/integrations/luminpdf)**

> **[Meetup MCP server](/docs/integrations/meetup)**

> **[MeetStream](/docs/integrations/meetstream)**

> **[Mem](/docs/integrations/mem)**

> **[Metaview](/docs/integrations/metaview)**

> **[Microsoft](/docs/integrations/microsoft)**

> **[Migma](/docs/integrations/migma)**

> **[Mixmax](/docs/integrations/mixmax)**

> **[MyFonts MCP server](/docs/integrations/monotype)**

> **[Novu](/docs/integrations/novu)**

> **[Otter](/docs/integrations/otter)**

> **[Outreach](/docs/integrations/outreach)**

> **[Pandadoc](/docs/integrations/pandadoc)**

> **[Paper](/docs/integrations/paper)**

> **[Pendo](/docs/integrations/pendo)**

> **[Plain](/docs/integrations/plain)**

> **[Quo](/docs/integrations/quo)**

> **[Readwise](/docs/integrations/readwise)**

> **[Refiner](/docs/integrations/refiner)**

> **[Rootly](/docs/integrations/rootly)**

> **[Send](/docs/integrations/send)**

> **[Shortcut](/docs/integrations/shortcut)**

> **[Storyblok](/docs/integrations/storyblok)**

> **[Streak](/docs/integrations/streak)**

> **[Superhuman Mail](/docs/integrations/superhuman)**

> **[Sybill](/docs/integrations/sybill)**

> **[Tally](/docs/integrations/tally)**

> **[Tana](/docs/integrations/tana)**

> **[Trustoo MCP Server](/docs/integrations/trustoo)**

> **[Wednesday.app MCP](/docs/integrations/wednesday)**

> **[Whimsical](/docs/integrations/whimsical)**

> **[Wispr Flow](/docs/integrations/wispr-flow)**

> **[Workato](/docs/integrations/workato)**

> **[Wix](/docs/integrations/wix)**

> **[Canva](/docs/integrations/canva)**

> **[Capacities](/docs/integrations/capacities)**

> **[Craft](/docs/integrations/craft)**

> **[Daily](/docs/integrations/daily)**

> **[Fireflies](/docs/integrations/fireflies)**

> **[Glean Developer Docs MCP Server](/docs/integrations/glean)**

> **[Goodnotes Remote MCP Server](/docs/integrations/goodnotes)**

> **[Google Mail](/docs/integrations/google-mail)**

> **[Granola](/docs/integrations/granola)**

> **[Grain](/docs/integrations/grain)**

> **[Gumloop](/docs/integrations/gumloop)**

> **[Help Scout Mailbox](/docs/integrations/helpscout-mailbox)**

> **[Jotform](/docs/integrations/jotform)**

> **[Lucid](/docs/integrations/lucid)**

> **[Make](/docs/integrations/make)**

> **[Miro](/docs/integrations/miro)**

> **[Mobbin](/docs/integrations/mobbin)**

> **[Prismic](/docs/integrations/prismic)**

> **[Read AI](/docs/integrations/read-ai)**

> **[Zapier](/docs/integrations/zapier)**

### Project management

> **[Asana](/docs/integrations/asana)**

> **[Basecamp](/docs/integrations/basecamp)**

> **[ClickUp](/docs/integrations/clickup)**

> **[Linear](/docs/integrations/linear)**

> **[Productboard](/docs/integrations/productboard)**

> **[Teamwork](/docs/integrations/teamwork)**

> **[TickTick](/docs/integrations/ticktick)**

> **[Todoist](/docs/integrations/todoist)**

> **[Trello](/docs/integrations/trello)**

### Developer tools

> **[Bitbucket](/docs/integrations/bitbucket)**

> **[Cloudinary](/docs/integrations/cloudinary)**

> **[DigitalOcean](/docs/integrations/digitalocean)**

> **[GitHub](/docs/integrations/github)**

> **[GitLab](/docs/integrations/gitlab)**

> **[incident.io](/docs/integrations/incident)**

> **[linkhut](/docs/integrations/linkhut)**

> **[PagerDuty](/docs/integrations/pagerduty)**

> **[Stack Exchange](/docs/integrations/stackexchange)**

> **[Vercel](/docs/integrations/vercel)**

> **[WakaTime](/docs/integrations/wakatime)**

> **[Agiflow](/docs/integrations/agiflow)**

> **[Alpic](/docs/integrations/alpic)**

> **[Anything](/docs/integrations/anything)**

> **[Apify](/docs/integrations/apify)**

> **[Atlassian](/docs/integrations/atlassian)**

> **[AuthZed MCP Server](/docs/integrations/authzed)**

> **[Autodesk](/docs/integrations/autodesk)**

> **[Avo](/docs/integrations/avo)**

> **[Background Removal](/docs/integrations/remove)**

> **[Blockscout](/docs/integrations/blockscout)**

> **[Braintrust](/docs/integrations/braintrust)**

> **[Builder Content](/docs/integrations/builder)**

> **[Cerebras Inference](/docs/integrations/cerebras)**

> **[Claude](/docs/integrations/claude)**

> **[Clerk](/docs/integrations/clerk)**

> **[Cloud-RF](/docs/integrations/cloudrf)**

> **[CockroachDB Cloud](/docs/integrations/cockroachlabs)**

> **[Cohere](/docs/integrations/cohere)**

> **[Context7 OAuth](/docs/integrations/context7)**

> **[CrewAI MCP Server](/docs/integrations/crewai)**

> **[Deepgram hosted MCP server](/docs/integrations/deepgram)**

> **[DevCycle](/docs/integrations/devcycle)**

> **[Domotz MCP Server (documented endpoint)](/docs/integrations/domotz)**

> **[Dremio Cloud](/docs/integrations/dremio)**

> **[Driftless](/docs/integrations/driftless)**

> **[Dub Partners](/docs/integrations/dub)**

> **[Exa](/docs/integrations/exa)**

> **[Executor](/docs/integrations/executor)**

> **[fal](/docs/integrations/fal)**

> **[Flagsmith](/docs/integrations/flagsmith)**

> **[Footprint](/docs/integrations/onefootprint)**

> **[Foxit PDF Editor MCP server](/docs/integrations/connectedpdf)**

> **[Framagit](/docs/integrations/framagit)**

> **[Glama MCP Inspector](/docs/integrations/glama)**

> **[GoDaddy MCP server](/docs/integrations/godaddy)**

> **[Google](/docs/integrations/google)**

> **[Google Sheet](/docs/integrations/google-sheet)**

> **[Grafana](/docs/integrations/grafana)**

> **[Gremlin](/docs/integrations/gremlin)**

> **[Hatchet Cloud](/docs/integrations/hatchet)**

> **[Hindsight Cloud MCP server (multi-bank)](/docs/integrations/vectorize)**

> **[hotglue](/docs/integrations/hotglue)**

> **[Huggingface](/docs/integrations/huggingface)**

> **[IP2Proxy Proxy Detection](/docs/integrations/ip2proxy)**

> **[Keploy Public](/docs/integrations/keploy)**

> **[Kestra MCP server](/docs/integrations/kestra)**

> **[Laminar](/docs/integrations/lmnr)**

> **[Langfuse](/docs/integrations/langfuse)**

> **[Linkly](/docs/integrations/linklyhq)**

> **[Manus](/docs/integrations/manus)**

> **[Mintlify](/docs/integrations/mintlify)**

> **[Mixedbread](/docs/integrations/mixedbread)**

> **[MongoDB Atlas App Services Admin](/docs/integrations/mongodb)**

> **[MotherDuck](/docs/integrations/motherduck)**

> **[Neurnav](/docs/integrations/neurnav)**

> **[Nuxt UI MCP Server](/docs/integrations/nuxt)**

> **[OpenRouter](/docs/integrations/openrouter)**

> **[OpenStatus](/docs/integrations/openstatus)**

> **[Package Search MCP Server](/docs/integrations/chroma)**

> **[Passage](/docs/integrations/passage)**

> **[Primitive](/docs/integrations/primitive)**

> **[Quicknode](/docs/integrations/quicknode)**

> **[Releases MCP server](/docs/integrations/releases)**

> **[Replicate](/docs/integrations/replicate)**

> **[ScrapingBee Remote MCP Server](/docs/integrations/scrapingbee)**

> **[ShareOut](/docs/integrations/shareout)**

> **[Skydive](/docs/integrations/skydive)**

> **[Smarty](/docs/integrations/smarty)**

> **[Stable Baseline](/docs/integrations/stablebaseline)**

> **[Statsig](/docs/integrations/statsig)**

> **[Statuspage](/docs/integrations/statuspage)**

> **[Tavily](/docs/integrations/tavily)**

> **[Tigris](/docs/integrations/tigrisdata)**

> **[UptimeRobot](/docs/integrations/uptimerobot)**

> **[Vectara REST API v2](/docs/integrations/vectara)**

> **[Venturu MCP server](/docs/integrations/venturu)**

> **[viaNexus vAST](/docs/integrations/blueskyapi)**

> **[Visualping](/docs/integrations/visualping)**

> **[W\&B](/docs/integrations/wandb)**

> **[Webhook Relay](/docs/integrations/webhookrelay)**

> **[Windmill](/docs/integrations/windmill)**

> **[Workos](/docs/integrations/workos)**

> **[Airbyte Agent](/docs/integrations/airbyte)**

> **[Axiom](/docs/integrations/axiom)**

> **[Backstage MCP server](/docs/integrations/backstage)**

> **[Bump.sh APIs hub MCP server](/docs/integrations/bump)**

> **[CircleCI REST API v2](/docs/integrations/circleci)**

> **[Cloudflare](/docs/integrations/cloudflare)**

> **[DatoCMS](/docs/integrations/datocms)**

> **[Docker Hub](/docs/integrations/docker)**

> **[Fern Docs MCP server](/docs/integrations/buildwithfern)**

> **[Gitpod Classic Public](/docs/integrations/gitpod)**

> **[GraphOS MCP Tools](/docs/integrations/apollographql)**

> **[Hex](/docs/integrations/hex)**

> **[Jira](/docs/integrations/jira)**

> **[Kensho LLM-Ready API](/docs/integrations/kensho)**

> **[Mermaid MCP Server](/docs/integrations/mermaid)**

> **[Monday](/docs/integrations/monday)**

> **[Neon](/docs/integrations/neon)**

> **[Netlify](/docs/integrations/netlify)**

> **[New Relic](/docs/integrations/newrelic)**

> **[pg-aiguide MCP server](/docs/integrations/tigerdata)**

> **[PlanetScale MCP server](/docs/integrations/pscale)**

> **[Pipedream](/docs/integrations/pipedream)**

> **[Prisma](/docs/integrations/prisma)**

> **[Pulumi](/docs/integrations/pulumi)**

> **[Qovery](/docs/integrations/qovery)**

> **[Redocly MCP Server](/docs/integrations/redocly)**

> **[Reducto](/docs/integrations/reducto)**

> **[Replit](/docs/integrations/replit-mcp)**

> **[Sanity](/docs/integrations/sanity)**

> **[Semaphore](/docs/integrations/semaphoreci)**

> **[Sentry](/docs/integrations/sentry)**

> **[SmartBear MCP Server (Swagger remote)](/docs/integrations/smartbear)**

> **[Supabase](/docs/integrations/supabase)**

> **[Transloadit hosted MCP server](/docs/integrations/transloadit)**

> **[Tuist](/docs/integrations/tuist)**

### CRM and sales

> **[Amplemarket](/docs/integrations/amplemarket)**

> **[Apollo](/docs/integrations/apollo)**

> **[Avoma](/docs/integrations/avoma)**

> **[Affinity](/docs/integrations/affinity)**

> **[Gong](/docs/integrations/gong)**

> **[Attio](/docs/integrations/attio)**

> **[HubSpot](/docs/integrations/hubspot)**

> **[Pipedrive](/docs/integrations/pipedrive)**

> **[Zoho Bigin](/docs/integrations/zoho-bigin)**

> **[Zoho CRM](/docs/integrations/zoho-crm)**

> **[Zoho Recruit](/docs/integrations/zoho-recruit)**

> **[HighLevel MCP (ChatGPT connector)](/docs/integrations/leadconnectorhq)**

> **[HubSpot REST APIs](/docs/integrations/hubapi)**

> **[OutCallerAI](/docs/integrations/outcallerai)**

### Support, marketing, and forms

> **[Acuity Scheduling](/docs/integrations/acuity-scheduling)**

> **[Cal.com](/docs/integrations/cal-com)**

> **[Calendly](/docs/integrations/calendly)**

> **[CleverReach](/docs/integrations/cleverreach)**

> **[Constant Contact](/docs/integrations/constant-contact)**

> **[Eventbrite](/docs/integrations/eventbrite)**

> **[Help Scout](/docs/integrations/helpscout)**

> **[Mailchimp](/docs/integrations/mailchimp)**

> **[Ubersuggest](/docs/integrations/ubersuggest)**

### Finance and commerce

> **[Agentcard](/docs/integrations/agentcard)**

> **[Carta](/docs/integrations/carta)**

> **[FreshBooks](/docs/integrations/freshbooks)**

> **[Harvest](/docs/integrations/harvest)**

> **[PayPal](/docs/integrations/paypal)**

> **[Printful](/docs/integrations/printful)**

> **[Ramp](/docs/integrations/ramp)**

> **[Splitwise](/docs/integrations/splitwise)**

> **[Square](/docs/integrations/square)**

> **[Xero](/docs/integrations/xero)**

> **[Zoho Books](/docs/integrations/zoho-books)**

> **[Zoho Inventory](/docs/integrations/zoho-inventory)**

> **[Zoho Invoice](/docs/integrations/zoho-invoice)**

> **[Alpaca](/docs/integrations/alpaca)**

> **[Benefits Intake](/docs/integrations/va)**

> **[Crypto.com Market Data MCP Server](/docs/integrations/crypto)**

> **[Dwolla](/docs/integrations/dwolla)**

> **[Increase](/docs/integrations/increase)**

> **[Kraken MCP server](/docs/integrations/kraken)**

> **[Lago](/docs/integrations/getlago)**

> **[Lemon Squeezy](/docs/integrations/lemonsqueezy)**

> **[Moov](/docs/integrations/moov)**

> **[Natural](/docs/integrations/natural)**

> **[NOWPayments](/docs/integrations/nowpayments)**

> **[OKX MCP chat endpoint](/docs/integrations/okx)**

> **[OpenBudget](/docs/integrations/openbudget)**

> **[Polar MCP server (production)](/docs/integrations/polar)**

> **[Rebilly](/docs/integrations/rebilly)**

> **[SecureLend AI Agents](/docs/integrations/securelend)**

> **[SubscriptionFlow MCP server](/docs/integrations/subscriptionflow)**

> **[YNAB](/docs/integrations/youneedabudget)**

> **[Airwallex AgentOS](/docs/integrations/airwallex)**

> **[Aiwyn Tax MCP Server](/docs/integrations/columnapi)**

> **[Brex](/docs/integrations/brex)**

> **[Gocardless](/docs/integrations/gocardless)**

> **[Mercury](/docs/integrations/mercury)**

> **[Optimizely Experimentation Data Authentication](/docs/integrations/optimizely)**

> **[SkyWatch](/docs/integrations/skywatch)**

> **[Spreedly Docs MCP Server](/docs/integrations/spreedly)**

> **[Stripe](/docs/integrations/stripe)**

### Zoho

> **[Zoho](/docs/integrations/zoho)**

> **[Zoho Calendar](/docs/integrations/zoho-calendar)**

> **[Zoho Desk](/docs/integrations/zoho-desk)**

> **[Zoho Mail](/docs/integrations/zoho-mail)**

> **[Zoho People](/docs/integrations/zoho-people)**

### Communication and social

> **[Ambient](/docs/integrations/ambient)**

> **[Archive](/docs/integrations/archive)**

> **[Bitly](/docs/integrations/bitly)**

> **[Buffer](/docs/integrations/buffer)**

> **[Discord](/docs/integrations/discord)**

> **[Outlook](/docs/integrations/outlook)**

> **[Spotify](/docs/integrations/spotify)**

> **[Timely](/docs/integrations/timely)**

> **[Tumblr](/docs/integrations/tumblr)**

> **[Twitch](/docs/integrations/twitch)**

> **[Vimeo](/docs/integrations/vimeo)**

> **[Webex](/docs/integrations/webex)**

> **[Webflow](/docs/integrations/webflow)**

> **[Windsor](/docs/integrations/windsor)**

> **[WordPress](/docs/integrations/wordpress)**

> **[Yahoo](/docs/integrations/yahoo)**

> **[Zoom](/docs/integrations/zoom)**

> **[Brevo](/docs/integrations/brevo)**

> **[CometChat MCP server](/docs/integrations/cometchat)**

> **[Cospark](/docs/integrations/cospark)**

> **[Knock](/docs/integrations/knock)**

> **[Loops](/docs/integrations/loops-so)**

> **[Mailerlite](/docs/integrations/mailerlite)**

> **[MailerSend](/docs/integrations/mailersend)**

> **[ManyChat](/docs/integrations/manychat)**

> **[Parker](/docs/integrations/parker)**

> **[Resend](/docs/integrations/resend)**

> **[Roam HQ](/docs/integrations/roam)**

> **[Sakari Messaging](/docs/integrations/sakari)**

> **[SIGNL4](/docs/integrations/signl4)**

> **[Slack](/docs/integrations/slack)**

> **[sms77.io](/docs/integrations/sms77)**

> **[SMTP2GO MCP server](/docs/integrations/smtp2go)**

> **[Typefully](/docs/integrations/typefully)**

> **[X](/docs/integrations/x)**

### Analytics

> **[Advanced Web Ranking](/docs/integrations/advancedwebranking)**

> **[Ahrefs](/docs/integrations/ahrefs)**

> **[Amplitude](/docs/integrations/amplitude)**

> **[CB Insights](/docs/integrations/cbinsights)**

> **[Clarity AI MCP server](/docs/integrations/clarity)**

> **[Databricks Workspace](/docs/integrations/databricks-workspace)**

> **[ClickHouse](/docs/integrations/clickhouse)**

> **[Contentsquare](/docs/integrations/contentsquare)**

> **[DataForSEO](/docs/integrations/dataforseo)**

> **[EODHD MCP Server v2](/docs/integrations/eodhd)**

> **[Explorium](/docs/integrations/explorium)**

> **[FMP](/docs/integrations/financialmodelingprep)**

> **[Fullstory](/docs/integrations/fullstory)**

> **[Local Falcon](/docs/integrations/localfalcon)**

> **[LogRocket](/docs/integrations/logrocket)**

> **[Massive](/docs/integrations/massive)**

> **[Morningstar](/docs/integrations/morningstar)**

> **[Motion Creative Analytics](/docs/integrations/motionapp)**

> **[Posthog](/docs/integrations/posthog)**

> **[SerpApi](/docs/integrations/serpapi)**

> **[Serpstat](/docs/integrations/serpstat)**

> **[SigNoz](/docs/integrations/signoz)**

> **[Snowflake](/docs/integrations/snowflake)**

> **[Supermetrics Marketing Analytics](/docs/integrations/supermetrics)**

> **[Synapse](/docs/integrations/synapse)**

> **[Ubersuggest MCP Server](/docs/integrations/neilpatelapi)**

> **[Vantage](/docs/integrations/vantage)**

### E-commerce

> **[Faire Wholesale MCP server](/docs/integrations/faire)**

> **[Helium 10](/docs/integrations/pacvue)**

> **[Imovirtual MCP Server](/docs/integrations/olx)**

> **[Kopi](/docs/integrations/trykopi)**

> **[Paddle](/docs/integrations/paddle)**

> **[Razorpay](/docs/integrations/razorpay)**

> **[ReCommerce MCP Server](/docs/integrations/recommerce)**

> **[Shop MCP server](/docs/integrations/shop)**

> **[Squareup](/docs/integrations/squareup)**

> **[Raptive Food MCP](/docs/integrations/raptive)**

### Travel and booking

> **[Jinko](/docs/integrations/gojinko)**

### HR and recruiting

> **[Ashby](/docs/integrations/ashby)**

> **[ApplyBoard MCP](/docs/integrations/applyboard)**

> **[Experteer MCP server](/docs/integrations/experteer)**

> **[Gusto](/docs/integrations/gusto)**

> **[LeapScholar MCP Server](/docs/integrations/leapscholar)**

> **[Magneto Empleos MCP Server](/docs/integrations/magneto365)**

> **[Spinify](/docs/integrations/spinify)**

### Health and wellness

> **[Habitify](/docs/integrations/habitify)**

### Media and entertainment

> **[Apple Music](/docs/integrations/apple)**

> **[Splice](/docs/integrations/splice)**

> **[VEED Fabric](/docs/integrations/veed)**

> **[GIPHY](/docs/integrations/giphy)**

### Security and compliance

> **[Arcjet](/docs/integrations/arcjet)**

> **[Flarehawk](/docs/integrations/flarehawk)**

> **[iubenda Consent Database](/docs/integrations/iubenda)**

> **[KYC Document API / Customer API v3.0](/docs/integrations/trulioo)**

> **[Malwarebytes](/docs/integrations/malwarebytes)**

> **[Midpage Legal Research](/docs/integrations/midpage)**

> **[Probely](/docs/integrations/probely)**

> **[Spectra Assure Community](/docs/integrations/reversinglabs)**

### Education

> **[Udemy Business](/docs/integrations/udemy)**

### More integrations

> **[LocationIQ](/docs/integrations/locationiq)**


# CB Insights
Source: https://getskydive.com/docs/integrations/cbinsights

Market intelligence on startups and private companies

## Example agent tasks

```text theme={null}
Connect CB Insights and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from CB Insights.
```

```text theme={null}
Use the agent to interact with CB Insights directly from your conversation.
```

## Connecting

Connect CB Insights over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add CB Insights.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach CB Insights through the MCP server.

## What you share

Connecting grants the agent access to CB Insights on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Cerebras Inference
Source: https://getskydive.com/docs/integrations/cerebras

High-speed AI inference on custom chips

## Example agent tasks

```text theme={null}
Connect Cerebras Inference and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Cerebras Inference.
```

```text theme={null}
Use the agent to interact with Cerebras Inference directly from your conversation.
```

## Connecting

Connect Cerebras Inference with an API key.

**1. Add the connection**

Generate an API key from your Cerebras Inference account settings.

**2. Authorize**

Copy the key. It is only shown once.

**3. The agent gets to work**

Paste it when prompted in the agent's connection card for Cerebras Inference.

## What you share

Connecting grants the agent access to Cerebras Inference on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Package Search MCP Server
Source: https://getskydive.com/docs/integrations/chroma

Search across public code package registries

## Example agent tasks

```text theme={null}
Connect Package Search MCP Server and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Package Search MCP Server.
```

```text theme={null}
Use the agent to interact with Package Search MCP Server directly from your conversation.
```

## Connecting

Connect Package Search over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Package Search.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Package Search through the MCP server.

## What you share

Connecting grants the agent access to Package Search MCP Server on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# CircleCI REST API v2
Source: https://getskydive.com/docs/integrations/circleci

CI/CD with the widest range of native integrations

## Example agent tasks

```text theme={null}
Connect CircleCI REST API v2 and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from CircleCI REST API v2.
```

```text theme={null}
Use the agent to interact with CircleCI REST API v2 directly from your conversation.
```

## Connecting

Connect CircleCI REST API v2 with an API key.

**1. Add the connection**

Generate an API key from your CircleCI REST API v2 account settings.

**2. Authorize**

Copy the key. It is only shown once.

**3. The agent gets to work**

Paste it when prompted in the agent's connection card for CircleCI REST API v2.

## What you share

Connecting grants the agent access to CircleCI REST API v2 on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Clarify
Source: https://getskydive.com/docs/integrations/clarify

AI-native customer relationship management

## Example agent tasks

```text theme={null}
Connect Clarify and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Clarify.
```

```text theme={null}
Use the agent to interact with Clarify directly from your conversation.
```

## Connecting

Connect Clarify over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Clarify.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Clarify through the MCP server.

## What you share

Connecting grants the agent access to Clarify on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Clarity AI MCP server
Source: https://getskydive.com/docs/integrations/clarity

Sustainability and ESG data for investors

## Example agent tasks

```text theme={null}
Connect Clarity AI MCP server and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Clarity AI MCP server.
```

```text theme={null}
Use the agent to interact with Clarity AI MCP server directly from your conversation.
```

## Connecting

Connect Clarity AI over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Clarity AI.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Clarity AI through the MCP server.

## What you share

Connecting grants the agent access to Clarity AI MCP server on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Claude
Source: https://getskydive.com/docs/integrations/claude

Call Claude AI models from your agent using the Anthropic API.

Claude is Anthropic's family of AI models. Connect it with an API key and your agents can send messages to any Claude model, including Haiku, Sonnet, and Opus, directly through the Anthropic Messages API.

## What the agent can do

> **Send messages to Claude**
>
> Call any Claude model with a prompt and get a response back. Works for summarization, classification, drafting, analysis, and any other text task.

> **Choose the right model**
>
> Discover available models with the models endpoint and pick the one that fits your task's cost and capability requirements.

> **Build multi-step pipelines**
>
> Chain Claude calls together with your other agent tools to handle complex workflows that need language understanding at multiple steps.

> **Process long context**
>
> Pass long documents, transcripts, or codebases into a Claude model and extract structured answers, edits, or summaries.

## Example agent tasks

```text theme={null}
Summarize this transcript and pull out the three most important action items.
```

```text theme={null}
Classify each of these support tickets as billing, technical, or general, then draft a reply for each one.
```

```text theme={null}
Review this pull request description and suggest improvements to the summary and testing notes.
```

## What you share

Connecting grants your agents access to the Anthropic Messages API using a workspace-scoped API key. The key reaches the API only; it does not give access to your claude.ai web or desktop account, chat history, or subscription. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Create a workspace-scoped API key**

Go to [console.anthropic.com/settings/keys](https://console.anthropic.com/settings/keys). Select a workspace from the dropdown before clicking Create Key. A workspace-scoped key is more predictable than an unscoped admin key and works cleanly with the proxy.

**2. Add the Claude connection**

From your agent's connections, choose Claude. Paste your API key when prompted.

**3. Your agents get to work**

They can now call Claude models through the Anthropic API. The platform injects the key as the `x-api-key` header automatically.

## Learn more

> **[Anthropic API docs](https://docs.anthropic.com/)**
>
> Full reference for the Claude models and Messages API.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Clerk
Source: https://getskydive.com/docs/integrations/clerk

User authentication and management for apps

## Example agent tasks

```text theme={null}
Connect Clerk and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Clerk.
```

```text theme={null}
Use the agent to interact with Clerk directly from your conversation.
```

## Connecting

Connect Clerk over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Clerk.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Clerk through the MCP server.

## What you share

Connecting grants the agent access to Clerk on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# CleverReach
Source: https://getskydive.com/docs/integrations/cleverreach

Let your agent work with your CleverReach email marketing

CleverReach is an email marketing platform for newsletters and campaigns. Connect it and your agent can work with your contacts, groups, and campaign data.

## What the agent can do

> **Manage contacts**
>
> Read and update subscribers in your groups.

> **Work with groups**
>
> See and organize your mailing lists.

> **Read campaigns**
>
> Pull campaign details and send data.

> **Check results**
>
> Review open, click, and bounce stats.

## Example agent tasks

```text theme={null}
After each send, report the open and click rates and flag what to try next.
```

```text theme={null}
When a contact comes in from a form, add them to the right CleverReach group.
```

```text theme={null}
Every Monday, tell me how the list grew last week and who bounced.
```

## What you share

During the connect flow you authorize what the agent can reach. It works only within what you grant. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Authorize CleverReach for the agent**

From the agent's connections, choose CleverReach and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[CleverReach developer docs](https://developers.cleverreach.com/)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# ClickHouse
Source: https://getskydive.com/docs/integrations/clickhouse

Let your agent query and analyze data in your ClickHouse Cloud database

ClickHouse is a columnar database built for real-time analytics. Connect it and your agent can run SQL queries, explore schemas, and surface analytical results from your ClickHouse Cloud data.

## What the agent can do

> **Run SQL queries**
>
> Execute analytical queries against your ClickHouse tables and return structured results.

> **Explore schemas**
>
> List databases, tables, and columns to understand the shape of your data before querying.

> **Aggregate and summarize**
>
> Run aggregations, group-by queries, and time-series rollups across large datasets.

> **Surface insights quickly**
>
> Take advantage of ClickHouse's columnar architecture to answer analytical questions fast.

## Example agent tasks

```text theme={null}
What are the top 10 events by volume in our events table over the last 7 days?
```

```text theme={null}
Show me the daily active users trend for this month, broken down by country.
```

```text theme={null}
List all tables in the analytics database and describe each one.
```

## What you share

During the connect flow you sign in to your ClickHouse Cloud account. The agent works within the permissions your database user allows. The token is stored securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Add the ClickHouse connection**

From the agent's connections, choose ClickHouse. The platform connects through ClickHouse Cloud's MCP server.

**2. Sign in to ClickHouse Cloud**

A connect card appears. Sign in to your ClickHouse Cloud account and approve access.

**3. The agent gets to work**

It can now run queries, explore your schema, and surface analytical results.

## Learn more

> **[ClickHouse docs](https://clickhouse.com/docs)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# ClickUp
Source: https://getskydive.com/docs/integrations/clickup

Let your agent manage your ClickUp tasks

ClickUp is a project management tool. Connect it and your agent can manage tasks, lists, docs, and time tracking across your workspace.

## Two ways to connect

ClickUp now runs a hosted MCP server. Connect it and the agent can work across your entire ClickUp workspace.

* **ClickUp MCP (recommended).** Connect ClickUp's MCP server once and the agent can create and update tasks, manage lists and folders, track time, post messages. See [Connect with MCP](#connect-with-mcp).
* **ClickUp OAuth.** The original sign-in integration, scoped to what you grant. See [Connect with OAuth](#connect-with-oauth).

## What the agent can do

> **Create and update tasks**
>
> Add tasks, change status, assignees, and due dates.

> **Browse the workspace**
>
> Navigate spaces, folders, and lists.

> **Track time**
>
> Start, stop, and log time entries on tasks.

> **Work with docs**
>
> Read and update pages inside ClickUp Docs.

## Example agent tasks

```text theme={null}
Create a task for each item in this email thread and assign them to the right people.
```

```text theme={null}
Tell me what is currently in review and what is blocked.
```

```text theme={null}
Log 2 hours against the design task I just finished.
```

## Connect with MCP

**1. Add the ClickUp connection**

From the agent's connections, choose ClickUp. The platform connects you through ClickUp's MCP server.

**2. Sign in to ClickUp**

A connect card appears. Sign in to your ClickUp account and approve.

**3. The agent gets to work**

It can now create and update tasks, manage lists and folders, track time, post messages through the MCP server.

> [!NOTE]
>
> MCP is the first-class connection when both options are available. See [Tools & MCP](/docs/capabilities/tools-and-mcp).

## Connect with OAuth

## What you share

During the connect flow you authorize what the agent can reach. It works only within what you grant. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

**1. Authorize ClickUp for the agent**

From the agent's connections, choose ClickUp and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[ClickUp developer docs](https://developer.clickup.com)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Cloudflare
Source: https://getskydive.com/docs/integrations/cloudflare

Content delivery, DNS and network security platform

## Example agent tasks

```text theme={null}
Connect Cloudflare and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Cloudflare.
```

```text theme={null}
Use the agent to interact with Cloudflare directly from your conversation.
```

## Connecting

Connect Cloudflare over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Cloudflare.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Cloudflare through the MCP server.

## What you share

Connecting grants the agent access to Cloudflare on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Cloudinary
Source: https://getskydive.com/docs/integrations/cloudinary

Let your agent manage and transform media assets in Cloudinary

Cloudinary is an image and video management platform. Connect it and your agent can upload media, apply transformations, organize asset libraries, and retrieve files from your Cloudinary account.

## What the agent can do

> **Upload and organize assets**
>
> Upload images and videos, set tags and metadata, and arrange files into folders.

> **Apply transformations**
>
> Resize, crop, convert formats, and apply effects or overlays on your media.

> **Search and retrieve**
>
> Query your asset library by tag, folder, format, or custom metadata.

> **Manage delivery settings**
>
> Update public IDs, access controls, and delivery profiles for your media files.

## Example agent tasks

```text theme={null}
Upload these product images to the /catalog folder and apply a 800x600 crop.
```

```text theme={null}
Find all assets tagged "draft" and move them to the archive folder.
```

```text theme={null}
Convert all PNGs in /hero to WebP and return the new delivery URLs.
```

## What you share

During the connect flow you sign in to your Cloudinary account. The agent works within what you grant. The token is stored securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Add the Cloudinary connection**

From the agent's connections, choose Cloudinary. The platform connects through Cloudinary's MCP server.

**2. Sign in to Cloudinary**

A connect card appears. Sign in to your Cloudinary account and approve access.

**3. The agent gets to work**

It can now upload, organize, transform, and retrieve assets from your media library.

## Learn more

> **[Cloudinary docs](https://cloudinary.com/developers)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Cloud-RF
Source: https://getskydive.com/docs/integrations/cloudrf

Radio signal propagation modeling and mapping

## Example agent tasks

```text theme={null}
Connect Cloud-RF and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Cloud-RF.
```

```text theme={null}
Use the agent to interact with Cloud-RF directly from your conversation.
```

## Connecting

Connect Cloud-RF with an API key.

**1. Add the connection**

Generate an API key from your Cloud-RF account settings.

**2. Authorize**

Copy the key. It is only shown once.

**3. The agent gets to work**

Paste it when prompted in the agent's connection card for Cloud-RF.

## What you share

Connecting grants the agent access to Cloud-RF on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# CockroachDB Cloud
Source: https://getskydive.com/docs/integrations/cockroachlabs

Distributed SQL database in the cloud

## Example agent tasks

```text theme={null}
Connect CockroachDB Cloud and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from CockroachDB Cloud.
```

```text theme={null}
Use the agent to interact with CockroachDB Cloud directly from your conversation.
```

## Connecting

Connect CockroachDB Cloud over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add CockroachDB Cloud.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach CockroachDB Cloud through the MCP server.

## What you share

Connecting grants the agent access to CockroachDB Cloud on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Cohere
Source: https://getskydive.com/docs/integrations/cohere

Large language models and AI API

## Example agent tasks

```text theme={null}
Connect Cohere and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Cohere.
```

```text theme={null}
Use the agent to interact with Cohere directly from your conversation.
```

## Connecting

Connect Cohere with an API key.

**1. Add the connection**

Generate an API key from your Cohere account settings.

**2. Authorize**

Copy the key. It is only shown once.

**3. The agent gets to work**

Paste it when prompted in the agent's connection card for Cohere.

## What you share

Connecting grants the agent access to Cohere on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Aiwyn Tax MCP Server
Source: https://getskydive.com/docs/integrations/columnapi

AI tax estimation from client tax documents

## Example agent tasks

```text theme={null}
Connect Aiwyn Tax MCP Server and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Aiwyn Tax MCP Server.
```

```text theme={null}
Use the agent to interact with Aiwyn Tax MCP Server directly from your conversation.
```

## Connecting

Connect Aiwyn Tax over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Aiwyn Tax.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Aiwyn Tax through the MCP server.

## What you share

Connecting grants the agent access to Aiwyn Tax MCP Server on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# CometChat MCP server
Source: https://getskydive.com/docs/integrations/cometchat

In-app chat and messaging SDK

## Example agent tasks

```text theme={null}
Connect CometChat MCP server and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from CometChat MCP server.
```

```text theme={null}
Use the agent to interact with CometChat MCP server directly from your conversation.
```

## Connecting

Connect CometChat over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add CometChat.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach CometChat through the MCP server.

## What you share

Connecting grants the agent access to CometChat MCP server on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Common Room
Source: https://getskydive.com/docs/integrations/commonroom

Go-to-market platform for buyer intelligence

## Example agent tasks

```text theme={null}
Connect Common Room and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Common Room.
```

```text theme={null}
Use the agent to interact with Common Room directly from your conversation.
```

## Connecting

Connect Common Room over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Common Room.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Common Room through the MCP server.

## What you share

Connecting grants the agent access to Common Room on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Foxit PDF Editor MCP server
Source: https://getskydive.com/docs/integrations/connectedpdf

PDF creation and editing software

## Example agent tasks

```text theme={null}
Connect Foxit PDF Editor MCP server and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Foxit PDF Editor MCP server.
```

```text theme={null}
Use the agent to interact with Foxit PDF Editor MCP server directly from your conversation.
```

## Connecting

Connect Foxit PDF Editor over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Foxit PDF Editor.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Foxit PDF Editor through the MCP server.

## What you share

Connecting grants the agent access to Foxit PDF Editor MCP server on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Constant Contact
Source: https://getskydive.com/docs/integrations/constant-contact

Let your agent work with your Constant Contact email marketing

Constant Contact runs email marketing and contacts. Connect it and your agent can read your contacts and campaign results.

## What the agent can do

> **Read contacts**
>
> See your contact lists.

> **Read campaigns**
>
> Pull campaign details and stats.

> **See results**
>
> Read open and click rates.

> **Add contacts**
>
> Add contacts to a list.

## Example agent tasks

```text theme={null}
After each send, report the results and flag what to change next time.
```

```text theme={null}
When a contact comes in from a form, add them to the right list automatically.
```

```text theme={null}
Every Monday, tell me how my list is growing and who's disengaging.
```

## What you share

During the connect flow you authorize what the agent can reach. It works only within what you grant. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Authorize Constant Contact for the agent**

From the agent's connections, choose Constant Contact and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[Constant Contact API docs](https://developer.constantcontact.com/api_guide/index.html)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Contentful
Source: https://getskydive.com/docs/integrations/contentful

Let your agent manage entries in your Contentful space

Contentful is a headless CMS. Connect it and your agent can read entries, publish content, and manage assets across your spaces.

## Two ways to connect

Contentful now runs a hosted MCP server with two regional endpoints: global (mcp.contentful.com) and EU (mcp.eu.contentful.com). The platform connects you to the global endpoint by default.

* **Contentful MCP (recommended).** Connect Contentful's MCP server once and the agent can read and publish entries, manage assets, inspect content types. See [Connect with MCP](#connect-with-mcp).
* **Contentful OAuth.** The original sign-in integration, scoped to what you grant. See [Connect with OAuth](#connect-with-oauth).

## What the agent can do

> **Read and search entries**
>
> Query content entries by type, field, or keyword.

> **Publish content**
>
> Publish and unpublish entries and assets.

> **Manage assets**
>
> List, update, and publish media assets.

> **Inspect content types**
>
> Read the schema that structures your content.

## Example agent tasks

```text theme={null}
Find all blog posts published this month and summarize the main topics.
```

```text theme={null}
Publish the draft entry for our new product page.
```

```text theme={null}
List the content types in my main space and describe each one.
```

## Connect with MCP

**1. Add the Contentful connection**

From the agent's connections, choose Contentful. The platform connects you through Contentful's MCP server.

**2. Sign in to Contentful**

A connect card appears. Sign in to your Contentful account and approve.

**3. The agent gets to work**

It can now read and publish entries, manage assets, inspect content types through the MCP server.

> [!NOTE]
>
> MCP is the first-class connection when both options are available. See [Tools & MCP](/docs/capabilities/tools-and-mcp).

## Connect with OAuth

## What you share

During the connect flow you authorize what the agent can reach. It works only within what you grant. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

**1. Authorize Contentful for the agent**

From the agent's connections, choose Contentful and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[Contentful developer docs](https://www.contentful.com/developers/docs)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Contentsquare
Source: https://getskydive.com/docs/integrations/contentsquare

Digital experience and behavior analytics

## Example agent tasks

```text theme={null}
Connect Contentsquare and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Contentsquare.
```

```text theme={null}
Use the agent to interact with Contentsquare directly from your conversation.
```

## Connecting

Connect Contentsquare over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Contentsquare.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Contentsquare through the MCP server.

## What you share

Connecting grants the agent access to Contentsquare on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Context7 OAuth
Source: https://getskydive.com/docs/integrations/context7

Up-to-date library documentation for AI coding

## Example agent tasks

```text theme={null}
Connect Context7 OAuth and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Context7 OAuth.
```

```text theme={null}
Use the agent to interact with Context7 OAuth directly from your conversation.
```

## Connecting

Connect Context7 OAuth over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Context7 OAuth.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Context7 OAuth through the MCP server.

## What you share

Connecting grants the agent access to Context7 OAuth on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Cosmic Hosted MCP (agent scope)
Source: https://getskydive.com/docs/integrations/cosmicjs

AI-native CMS with agents that build and deploy content

## Example agent tasks

```text theme={null}
Connect Cosmic Hosted MCP (agent scope) and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Cosmic Hosted MCP (agent scope).
```

```text theme={null}
Use the agent to interact with Cosmic Hosted MCP (agent scope) directly from your conversation.
```

## Connecting

Connect Cosmic Hosted (agent scope) over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Cosmic Hosted MCP (agent scope).

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Cosmic Hosted MCP (agent scope) through the MCP server.

## What you share

Connecting grants the agent access to Cosmic Hosted MCP (agent scope) on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Cospark
Source: https://getskydive.com/docs/integrations/cospark

AI ad maker for video and image ads - research, script, and produce creative at scale

Cospark is an AI ad maker that takes your product, brief, script, or footage and turns it into a finished video or image ad. Give it a sentence and it researches the angle, writes the script, generates footage, and edits everything into a ready-to-launch cut your agents can request or iterate on.

## What your agents can do

> **Create video ads**
>
> Generate UGC-style, talking-head, product, voiceover, and animated video ads from a brief or product description.

> **Script and produce from scratch**
>
> Research your product, write a hook and script, then produce a fully edited first cut without manual prompting.

> **Iterate on existing ads**
>
> Refine a timeline by chat: cut clips, adjust pacing, swap voiceover, or rework captions until the ad is right.

> **Reference competitor ads**
>
> Paste a Facebook or Instagram ad as a reference and get a new version structured for your product and brand voice.

## Example agent tasks

```text theme={null}
Create a 30-second UGC-style video ad for our new protein powder using our brand guidelines.
```

```text theme={null}
Take this competitor Facebook ad as a reference and produce a version for our running shoes.
```

```text theme={null}
Trim the talking-head ad to 15 seconds and add captions synced to the voiceover.
```

## What you share

Connecting gives your agents access to your Cospark workspace: your brand guidelines, uploaded products, visual identity, and any ad projects in your account. Agents can create new ad projects, generate footage, and edit timelines. Requires an active Cospark account. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Add the Cospark connection**

From the agent's connections, choose Cospark. The platform connects through Cospark's MCP server.

**2. Sign in to Cospark**

A connect card appears. Sign in with your Cospark account at cospark.so and approve access.

**3. Your agents get to work**

They can now create, generate, and iterate on video and image ad creative directly from your Cospark workspace.

## Learn more

> **[Cospark](https://cospark.so)**
>
> Cospark's AI ad maker and product overview.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Craft
Source: https://getskydive.com/docs/integrations/craft

Docs and notes app for teams

## Example agent tasks

```text theme={null}
Connect Craft and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Craft.
```

```text theme={null}
Use the agent to interact with Craft directly from your conversation.
```

## Connecting

Connect Craft over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Craft.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Craft through the MCP server.

## What you share

Connecting grants the agent access to Craft on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# CrewAI MCP Server
Source: https://getskydive.com/docs/integrations/crewai

Framework for orchestrating multi-agent AI workflows

## Example agent tasks

```text theme={null}
Connect CrewAI MCP Server and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from CrewAI MCP Server.
```

```text theme={null}
Use the agent to interact with CrewAI MCP Server directly from your conversation.
```

## Connecting

Connect CrewAI over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add CrewAI.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach CrewAI through the MCP server.

## What you share

Connecting grants the agent access to CrewAI MCP Server on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Crossbeam
Source: https://getskydive.com/docs/integrations/crossbeam

Partner ecosystem data sharing for sales teams

## Example agent tasks

```text theme={null}
Connect Crossbeam and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Crossbeam.
```

```text theme={null}
Use the agent to interact with Crossbeam directly from your conversation.
```

## Connecting

Connect Crossbeam over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Crossbeam.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Crossbeam through the MCP server.

## What you share

Connecting grants the agent access to Crossbeam on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Crypto.com Market Data MCP Server
Source: https://getskydive.com/docs/integrations/crypto

Cryptocurrency price and market data

## Example agent tasks

```text theme={null}
Connect Crypto.com Market Data MCP Server and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Crypto.com Market Data MCP Server.
```

```text theme={null}
Use the agent to interact with Crypto.com Market Data MCP Server directly from your conversation.
```

## Connecting

Connect Crypto.com Market Data over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Crypto.com Market Data.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Crypto.com Market Data through the MCP server.

## What you share

Connecting grants the agent access to Crypto.com Market Data MCP Server on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Cube
Source: https://getskydive.com/docs/integrations/cubesoftware

Financial planning and analysis software

## Example agent tasks

```text theme={null}
Connect Cube and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Cube.
```

```text theme={null}
Use the agent to interact with Cube directly from your conversation.
```

## Connecting

Connect Cube over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Cube.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Cube through the MCP server.

## What you share

Connecting grants the agent access to Cube on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Daily
Source: https://getskydive.com/docs/integrations/daily

Video calling and WebRTC infrastructure API

## Example agent tasks

```text theme={null}
Connect Daily and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Daily.
```

```text theme={null}
Use the agent to interact with Daily directly from your conversation.
```

## Connecting

Connect Daily with an API key.

**1. Add the connection**

Generate an API key from your Daily account settings.

**2. Authorize**

Copy the key. It is only shown once.

**3. The agent gets to work**

Paste it when prompted in the agent's connection card for Daily.

## What you share

Connecting grants the agent access to Daily on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Databricks Workspace
Source: https://getskydive.com/docs/integrations/databricks-workspace

Unified analytics workspace for data engineering, notebooks, and clusters

Databricks Workspace is Databricks' unified analytics workspace. Connect it with a workspace personal access token and your agents can inspect clusters, jobs, SQL warehouses, and Unity Catalog through the Databricks REST API.

## What your agents can do

> **Inspect clusters and jobs**
>
> List clusters, jobs, and SQL warehouses so agents can see what is running in your workspace before they propose changes.

> **Explore Unity Catalog**
>
> Walk catalogs, schemas, and tables so agents can map your data assets before writing queries or jobs.

> **Report on workspace health**
>
> Summarize cluster utilization, stuck jobs, or warehouse load in plain language for the team.

> **Prepare write actions**
>
> Draft job configs or cluster changes and run them only after you confirm, since writes are always confirmed first.

## Example agent tasks

```text theme={null}
List the clusters in our Databricks workspace and flag any that have been running longer than 24 hours.
```

```text theme={null}
Walk the Unity Catalog and describe what each schema in the analytics catalog contains.
```

```text theme={null}
Summarize yesterday's job failures from our workspace and draft a short incident note.
```

## What you share

Connecting grants your agents access to one Databricks workspace through the REST API (api/2.0) using a workspace personal access token. Reads are the default; agents confirm with you before creating or running clusters or jobs, or executing SQL. Tokens are workspace-scoped and auto-revoke after 90 days of disuse; if a call 401s, generate a fresh token. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Generate a workspace access token**

In your Databricks workspace, click your username, then Settings > Developer > Access tokens > Manage > Generate new token. The token is shown once; copy it. Also have your workspace URL ready (for example `https://dbc-<name>.cloud.databricks.com`, or the Azure or GCP equivalent).

**2. Add the Databricks Workspace connection**

From your agent's connections, choose Databricks Workspace. Paste your token and workspace URL when prompted. The platform injects the token as the `Authorization: Bearer` header automatically.

**3. Your agents get to work**

They can now inspect and work in your workspace through the Databricks REST API.

## Learn more

> **[Databricks REST API reference](https://docs.databricks.com/aws/en/dev-tools/api/latest/)**
>
> Full reference for the REST API endpoints your agents use.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# DataForSEO
Source: https://getskydive.com/docs/integrations/dataforseo

SEO and search engine results data API

## Example agent tasks

```text theme={null}
Connect DataForSEO and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from DataForSEO.
```

```text theme={null}
Use the agent to interact with DataForSEO directly from your conversation.
```

## Connecting

Connect DataForSEO over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add DataForSEO.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach DataForSEO through the MCP server.

## What you share

Connecting grants the agent access to DataForSEO on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# DatoCMS
Source: https://getskydive.com/docs/integrations/datocms

GraphQL-first headless CMS with image processing

## Example agent tasks

```text theme={null}
Connect DatoCMS and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from DatoCMS.
```

```text theme={null}
Use the agent to interact with DatoCMS directly from your conversation.
```

## Connecting

**MCP server**

Connect DatoCMS over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add DatoCMS.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach DatoCMS through the MCP server.

**API key**

Connect Content Delivery with an API key.

**1. Add the connection**

Generate an API key from your DatoCMS account settings.

**2. Authorize**

Copy the key. It is only shown once.

**3. The agent gets to work**

Paste it when prompted in the agent's connection card for DatoCMS.

## What you share

Connecting grants the agent access to DatoCMS on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Day AI
Source: https://getskydive.com/docs/integrations/day

AI customer memory and sales agents

## Example agent tasks

```text theme={null}
Connect Day AI and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Day AI.
```

```text theme={null}
Use the agent to interact with Day AI directly from your conversation.
```

## Connecting

Connect Day AI over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Day AI.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Day AI through the MCP server.

## What you share

Connecting grants the agent access to Day AI on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Deepgram hosted MCP server
Source: https://getskydive.com/docs/integrations/deepgram

Speech-to-text and voice AI transcription

## Example agent tasks

```text theme={null}
Connect Deepgram hosted MCP server and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Deepgram hosted MCP server.
```

```text theme={null}
Use the agent to interact with Deepgram hosted MCP server directly from your conversation.
```

## Connecting

Connect Deepgram hosted over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Deepgram hosted.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Deepgram hosted through the MCP server.

## What you share

Connecting grants the agent access to Deepgram hosted MCP server on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# DevCycle
Source: https://getskydive.com/docs/integrations/devcycle

Feature flag and experimentation platform

## Example agent tasks

```text theme={null}
Connect DevCycle and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from DevCycle.
```

```text theme={null}
Use the agent to interact with DevCycle directly from your conversation.
```

## Connecting

Connect DevCycle over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add DevCycle.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach DevCycle through the MCP server.

## What you share

Connecting grants the agent access to DevCycle on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# DigitalOcean
Source: https://getskydive.com/docs/integrations/digitalocean

Let your agent work with your DigitalOcean infrastructure

DigitalOcean hosts your servers, databases, and apps. Connect it and your agent can read your resources and their status.

## What the agent can do

> **Read Droplets**
>
> See your servers and their status.

> **See databases**
>
> Read managed database details.

> **Check status**
>
> Pull resource health and usage.

> **Find resources**
>
> Look up a specific resource.

## Example agent tasks

```text theme={null}
What's the status of my Droplets right now?
```

```text theme={null}
Every morning, flag any resource that looks unhealthy.
```

```text theme={null}
Summarize my current DigitalOcean footprint and costs.
```

## What you share

During the connect flow you authorize what the agent can reach. It works only within what you grant. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Authorize DigitalOcean for the agent**

From the agent's connections, choose DigitalOcean and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[DigitalOcean API docs](https://docs.digitalocean.com/reference/api/)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Discord
Source: https://getskydive.com/docs/integrations/discord

Let your agent read and post in your Discord servers

Discord runs community and team chat. Connect it and your agent can read messages in the channels you share and post to them.

## What the agent can do

> **Read messages**
>
> Catch up on a channel.

> **Post messages**
>
> Send updates or replies.

> **See channels**
>
> Find the right channel.

> **See members**
>
> Read who's in a server.

## Example agent tasks

```text theme={null}
Every morning, summarize what happened in our community overnight and flag what needs a reply.
```

```text theme={null}
When someone asks a common question, post the answer and tag me if it's new.
```

```text theme={null}
Watch for anything urgent in #support and DM me right away.
```

## What you share

During the connect flow you authorize what the agent can reach. It works only within what you grant. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Authorize Discord for the agent**

From the agent's connections, choose Discord and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[Discord API docs](https://discord.com/developers/docs/intro)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Docker Hub
Source: https://getskydive.com/docs/integrations/docker

Container image registry for developers

## Example agent tasks

```text theme={null}
Connect Docker Hub and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Docker Hub.
```

```text theme={null}
Use the agent to interact with Docker Hub directly from your conversation.
```

## Connecting

Connect Docker Hub with an API key.

**1. Add the connection**

Generate an API key from your Docker Hub account settings.

**2. Authorize**

Copy the key. It is only shown once.

**3. The agent gets to work**

Paste it when prompted in the agent's connection card for Docker Hub.

## What you share

Connecting grants the agent access to Docker Hub on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Domotz MCP Server (documented endpoint)
Source: https://getskydive.com/docs/integrations/domotz

Remote network monitoring for IT and MSPs

## Example agent tasks

```text theme={null}
Connect Domotz MCP Server (documented endpoint) and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Domotz MCP Server (documented endpoint).
```

```text theme={null}
Use the agent to interact with Domotz MCP Server (documented endpoint) directly from your conversation.
```

## Connecting

Connect Domotz (documented endpoint) over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Domotz MCP Server (documented endpoint).

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Domotz MCP Server (documented endpoint) through the MCP server.

## What you share

Connecting grants the agent access to Domotz MCP Server (documented endpoint) on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Dremio Cloud
Source: https://getskydive.com/docs/integrations/dremio

Cloud data lakehouse query engine

## Example agent tasks

```text theme={null}
Connect Dremio Cloud and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Dremio Cloud.
```

```text theme={null}
Use the agent to interact with Dremio Cloud directly from your conversation.
```

## Connecting

Connect Dremio Cloud with an API key.

**1. Add the connection**

Generate an API key from your Dremio Cloud account settings.

**2. Authorize**

Copy the key. It is only shown once.

**3. The agent gets to work**

Paste it when prompted in the agent's connection card for Dremio Cloud.

## What you share

Connecting grants the agent access to Dremio Cloud on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Driftless
Source: https://getskydive.com/docs/integrations/driftless

Shared codebase context for engineering teams and AI agents

## Example agent tasks

```text theme={null}
Connect Driftless and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Driftless.
```

```text theme={null}
Use the agent to interact with Driftless directly from your conversation.
```

## Connecting

Connect Driftless over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Driftless.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Driftless through the MCP server.

## What you share

Connecting grants the agent access to Driftless on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Dropbox
Source: https://getskydive.com/docs/integrations/dropbox

Let your agent work with your Dropbox files

Dropbox keeps your files synced in the cloud. Connect it and your agent can find, read, and save files.

## What the agent can do

> **Find files**
>
> Search your Dropbox.

> **Read files**
>
> Pull content from a file.

> **Save files**
>
> Upload new files to a folder.

> **Browse folders**
>
> Navigate your folders.

## Example agent tasks

```text theme={null}
When a client uploads assets, sort them into the right project folder automatically.
```

```text theme={null}
Each Friday, package the week's deliverables into one shared folder and send me the link.
```

```text theme={null}
Find the latest version of every file with "final" in the name and tell me which is actually newest.
```

## What you share

During the connect flow you authorize what the agent can reach. It works only within what you grant. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Authorize Dropbox for the agent**

From the agent's connections, choose Dropbox and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[Dropbox API docs](https://www.dropbox.com/developers/documentation/http/documentation)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Dropbox Sign
Source: https://getskydive.com/docs/integrations/dropbox-sign

Let your agent work with e-signature requests

Dropbox Sign (formerly HelloSign) handles e-signatures. Connect it and your agent can read signature requests and their status.

## What the agent can do

> **Read requests**
>
> See your signature requests and who's signed.

> **Check status**
>
> Track what's pending and complete.

> **Send for signature**
>
> Start a signature request from a document.

> **Find a request**
>
> Locate a specific request.

## Example agent tasks

```text theme={null}
The moment a contract is fully signed, log it and kick off the onboarding checklist.
```

```text theme={null}
Every morning, tell me which signature requests have gone quiet for more than two days.
```

```text theme={null}
Send this agreement for signature and ping me when it's done.
```

## What you share

During the connect flow you authorize what the agent can reach. It works only within what you grant. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Authorize Dropbox Sign for the agent**

From the agent's connections, choose Dropbox Sign and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[Dropbox Sign API docs](https://developers.hellosign.com/docs)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Dub Partners
Source: https://getskydive.com/docs/integrations/dub

Link management and short URLs for partner programs

## Example agent tasks

```text theme={null}
Connect Dub Partners and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Dub Partners.
```

```text theme={null}
Use the agent to interact with Dub Partners directly from your conversation.
```

## Connecting

Connect Dub Partners over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Dub Partners.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Dub Partners through the MCP server.

## What you share

Connecting grants the agent access to Dub Partners on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Dwolla
Source: https://getskydive.com/docs/integrations/dwolla

ACH bank payment infrastructure

## Example agent tasks

```text theme={null}
Connect Dwolla and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Dwolla.
```

```text theme={null}
Use the agent to interact with Dwolla directly from your conversation.
```

## Connecting

Connect Dwolla with an API key.

**1. Add the connection**

Generate an API key from your Dwolla account settings.

**2. Authorize**

Copy the key. It is only shown once.

**3. The agent gets to work**

Paste it when prompted in the agent's connection card for Dwolla.

## What you share

Connecting grants the agent access to Dwolla on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Egnyte
Source: https://getskydive.com/docs/integrations/egnyte

Cloud content collaboration and file storage

## What the agent can do

> **Work with Egnyte**
>
> Egnyte gates MCP access at the tenant level, on the customer's side.

## Example agent tasks

```text theme={null}
Connect Egnyte and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Egnyte.
```

```text theme={null}
Use the agent to interact with Egnyte directly from your conversation.
```

## Connecting

Connect Egnyte over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Egnyte.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Egnyte through the MCP server.

## What you share

Connecting grants the agent access to Egnyte on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# EODHD MCP Server v2
Source: https://getskydive.com/docs/integrations/eodhd

Historical and real-time financial market data

## Example agent tasks

```text theme={null}
Connect EODHD MCP Server v2 and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from EODHD MCP Server v2.
```

```text theme={null}
Use the agent to interact with EODHD MCP Server v2 directly from your conversation.
```

## Connecting

Connect EODHD v1 over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add EODHD MCP Server v2.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach EODHD MCP Server v2 through the MCP server.

## What you share

Connecting grants the agent access to EODHD MCP Server v2 on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Eventbrite
Source: https://getskydive.com/docs/integrations/eventbrite

Let your agent work with your Eventbrite events

Eventbrite runs event ticketing and registration. Connect it and your agent can read your events and attendees.

## What the agent can do

> **Read events**
>
> See your events and details.

> **Read attendees**
>
> Pull who's registered.

> **See orders**
>
> Read ticket sales and orders.

> **Find an event**
>
> Locate the right event.

## Example agent tasks

```text theme={null}
As registrations come in, tell me when we're pacing behind and need a push.
```

```text theme={null}
The day before the event, send me the final headcount and the attendee list.
```

```text theme={null}
After the event, summarize turnout and who to follow up with.
```

## What you share

During the connect flow you authorize what the agent can reach. It works only within what you grant. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Authorize Eventbrite for the agent**

From the agent's connections, choose Eventbrite and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[Eventbrite API docs](https://www.eventbrite.com/platform/api)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Exa
Source: https://getskydive.com/docs/integrations/exa

AI-native search engine and web data API

## Example agent tasks

```text theme={null}
Connect Exa and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Exa.
```

```text theme={null}
Use the agent to interact with Exa directly from your conversation.
```

## Connecting

Connect Exa over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Exa.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Exa through the MCP server.

## What you share

Connecting grants the agent access to Exa on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Executor
Source: https://getskydive.com/docs/integrations/executor

Integration gateway connecting AI agents to apps

## Example agent tasks

```text theme={null}
Connect Executor and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Executor.
```

```text theme={null}
Use the agent to interact with Executor directly from your conversation.
```

## Connecting

Connect Executor over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Executor.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Executor through the MCP server.

## What you share

Connecting grants the agent access to Executor on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Experteer MCP server
Source: https://getskydive.com/docs/integrations/experteer

Executive job search and headhunting

## What the agent can do

> **Search Jobs**
>
> Use this tool to search for executive and specialist-level job opportunities listed on Experteer.de.

## Example agent tasks

```text theme={null}
Search Experteer MCP server for relevant results and summarize what you find.
```

```text theme={null}
Ask the agent to help you with Experteer MCP server tasks using natural language.
```

## Connecting

Connect Experteer over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Experteer.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Experteer through the MCP server.

## What you share

Connecting grants the agent access to Experteer MCP server on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Explorium
Source: https://getskydive.com/docs/integrations/explorium

B2B data enrichment for go-to-market teams

## Example agent tasks

```text theme={null}
Connect Explorium and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Explorium.
```

```text theme={null}
Use the agent to interact with Explorium directly from your conversation.
```

## Connecting

Connect Explorium over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Explorium.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Explorium through the MCP server.

## What you share

Connecting grants the agent access to Explorium on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Faire Wholesale MCP server
Source: https://getskydive.com/docs/integrations/faire

Wholesale marketplace for independent retailers

## Example agent tasks

```text theme={null}
Connect Faire Wholesale MCP server and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Faire Wholesale MCP server.
```

```text theme={null}
Use the agent to interact with Faire Wholesale MCP server directly from your conversation.
```

## Connecting

Connect Faire Wholesale over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Faire Wholesale.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Faire Wholesale through the MCP server.

## What you share

Connecting grants the agent access to Faire Wholesale MCP server on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# fal
Source: https://getskydive.com/docs/integrations/fal

Hosted generative AI model inference (FLUX and more)

fal hosts generative AI models for image, video, and audio generation. Your agents can search the model catalog, read model schemas and pricing, and submit generation jobs without leaving the conversation.

## What your agents can do

> **Search the model catalog**
>
> Find models by task and use recommend\_model when you know what you want but not which model fits.

> **Read schemas before running**
>
> get\_model\_schema returns each model's inputs, so the agent builds a valid request the first time.

> **Submit and track jobs**
>
> submit\_job starts a generation, check\_job reports its status until the output is ready.

> **Estimate cost**
>
> get\_pricing tells you what a run consumes in fal credits before the agent spends them.

## Example agent tasks

```text theme={null}
Find a FLUX model for generating product photos, check its schema and pricing, then generate three images.
```

```text theme={null}
List the video models on fal and estimate the credit cost of a 5-second clip for each.
```

```text theme={null}
Generate an image with my preferred fal model and check on the job until it finishes.
```

## Connecting

**MCP server**

Use fal hosted image, video and audio models through its official MCP server.

**1. Enable MCP access**

In your fal account, enable MCP access at fal.ai/dashboard/account/details first.

**2. Add the connection**

From your agent's tools, add fal and sign in with your fal account.

**3. The agent gets to work**

The agent reaches fal through the MCP server. Start with search\_models or recommend\_model, and read get\_model\_schema before run\_model.

**API key**

Call fal model APIs directly with a fal API key. No OAuth or browser sign-in needed.

**1. Create an API key**

Create one at [dashboard.fal.ai/keys](https://dashboard.fal.ai/keys).

**2. Add the connection**

From your agent's tools, add fal and paste your fal API key.

**3. The agent gets to work**

The agent calls the fal REST API (fal.run) with your key. Reading model docs or getting a schema via the MCP method first avoids malformed requests.

Generation consumes your fal credits.

## What you share

Connecting grants your agents access to fal on your behalf. With the API key method, the key authorizes model inference on your account. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

## Learn more

> **[fal documentation](https://fal.ai/docs/documentation)**
>
> Model catalog, pricing, and API reference for fal.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Fathom
Source: https://getskydive.com/docs/integrations/fathom

AI meeting notetaker and transcription

## Example agent tasks

```text theme={null}
Connect Fathom and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Fathom.
```

```text theme={null}
Use the agent to interact with Fathom directly from your conversation.
```

## Connecting

Connect Fathom over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Fathom.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Fathom through the MCP server.

## What you share

Connecting grants the agent access to Fathom on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Figma
Source: https://getskydive.com/docs/integrations/figma

Let your agent read your Figma files and designs

Figma is where design happens. Connect it and your agent can read your files, frames, and comments.

## What the agent can do

> **Read files**
>
> Pull the structure and content of a design file.

> **Read comments**
>
> See feedback left on a design.

> **Export frames**
>
> Get images of specific frames.

> **Find a file**
>
> Locate the right design file.

## Example agent tasks

```text theme={null}
Summarize the open comments across our design files so the team knows what to address.
```

```text theme={null}
When a design is marked ready, pull the specs into a ticket for engineering.
```

```text theme={null}
Tell me which screens changed in the latest revision of the mobile flow.
```

## What you share

During the connect flow you authorize what the agent can reach. It works only within what you grant. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Authorize Figma for the agent**

From the agent's connections, choose Figma and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[Figma API docs](https://www.figma.com/developers/api)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# FMP
Source: https://getskydive.com/docs/integrations/financialmodelingprep

Stock market and financial data API

## Example agent tasks

```text theme={null}
Connect FMP and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from FMP.
```

```text theme={null}
Use the agent to interact with FMP directly from your conversation.
```

## Connecting

Connect FMP with an API key.

**1. Add the connection**

Generate an API key from your FMP account settings.

**2. Authorize**

Copy the key. It is only shown once.

**3. The agent gets to work**

Paste it when prompted in the agent's connection card for FMP.

## What you share

Connecting grants the agent access to FMP on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Fireflies
Source: https://getskydive.com/docs/integrations/fireflies

Meeting transcription and notes

## Example agent tasks

```text theme={null}
Connect Fireflies and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Fireflies.
```

```text theme={null}
Use the agent to interact with Fireflies directly from your conversation.
```

## Connecting

**MCP server**

Connect Fireflies over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Fireflies.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Fireflies through the MCP server.

**API key**

Connect Fireflies with a bearer API key.

**1. Get an API key**

In Fireflies, open Integrations, then Fireflies API, and create or copy your key.

**2. Add the connection**

From the agent's connections, choose Fireflies and paste your API key when prompted.

**3. The agent gets to work**

The agent can now reach the Fireflies API on your behalf.

## What you share

Connecting grants the agent access to Fireflies on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Flagsmith
Source: https://getskydive.com/docs/integrations/flagsmith

Feature flag and remote config management

## Example agent tasks

```text theme={null}
Connect Flagsmith and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Flagsmith.
```

```text theme={null}
Use the agent to interact with Flagsmith directly from your conversation.
```

## Connecting

Connect Flagsmith over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Flagsmith.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Flagsmith through the MCP server.

## What you share

Connecting grants the agent access to Flagsmith on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Flarehawk
Source: https://getskydive.com/docs/integrations/flarehawk

Autonomous cloud security operations platform

## Example agent tasks

```text theme={null}
Connect Flarehawk and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Flarehawk.
```

```text theme={null}
Use the agent to interact with Flarehawk directly from your conversation.
```

## Connecting

Connect Flarehawk over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Flarehawk.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Flarehawk through the MCP server.

## What you share

Connecting grants the agent access to Flarehawk on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Framagit
Source: https://getskydive.com/docs/integrations/framagit

French nonprofit-run GitLab hosting

## Example agent tasks

```text theme={null}
Connect Framagit and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Framagit.
```

```text theme={null}
Use the agent to interact with Framagit directly from your conversation.
```

## Connecting

Connect Framagit with an API key.

**1. Add the connection**

Generate an API key from your Framagit account settings.

**2. Authorize**

Copy the key. It is only shown once.

**3. The agent gets to work**

Paste it when prompted in the agent's connection card for Framagit.

## What you share

Connecting grants the agent access to Framagit on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# FreshBooks
Source: https://getskydive.com/docs/integrations/freshbooks

Let your agent work with your FreshBooks accounting

FreshBooks handles invoicing and accounting for small businesses. Connect it and your agent can read invoices and clients, and create records.

## What the agent can do

> **Read invoices**
>
> See invoices and their status.

> **Read clients**
>
> Pull client details.

> **Create invoices**
>
> Add new invoices.

> **Track time**
>
> Read time entries and expenses.

## Example agent tasks

```text theme={null}
When a project wraps, generate the invoice from the tracked time and send it.
```

```text theme={null}
Every Monday, tell me what's outstanding and who's overdue.
```

```text theme={null}
Summarize what I billed this month versus last.
```

## What you share

During the connect flow you authorize what the agent can reach. It works only within what you grant. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Authorize FreshBooks for the agent**

From the agent's connections, choose FreshBooks and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[FreshBooks API docs](https://www.freshbooks.com/api/start)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Fullstory
Source: https://getskydive.com/docs/integrations/fullstory

Session replay and digital experience analytics

## Example agent tasks

```text theme={null}
Connect Fullstory and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Fullstory.
```

```text theme={null}
Use the agent to interact with Fullstory directly from your conversation.
```

## Connecting

Connect Fullstory over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Fullstory.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Fullstory through the MCP server.

## What you share

Connecting grants the agent access to Fullstory on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Garchi CMS MCP OAuth Endpoint
Source: https://getskydive.com/docs/integrations/garchi

Agent-native headless CMS

## Example agent tasks

```text theme={null}
Connect Garchi CMS MCP OAuth Endpoint and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Garchi CMS MCP OAuth Endpoint.
```

```text theme={null}
Use the agent to interact with Garchi CMS MCP OAuth Endpoint directly from your conversation.
```

## Connecting

Connect Garchi over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Garchi CMS MCP OAuth Endpoint.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Garchi CMS MCP OAuth Endpoint through the MCP server.

## What you share

Connecting grants the agent access to Garchi CMS MCP OAuth Endpoint on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Guru MCP Server
Source: https://getskydive.com/docs/integrations/getguru

Knowledge base and wiki for team answers

## Example agent tasks

```text theme={null}
Connect Guru MCP Server and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Guru MCP Server.
```

```text theme={null}
Use the agent to interact with Guru MCP Server directly from your conversation.
```

## Connecting

Connect Guru over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Guru.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Guru through the MCP server.

## What you share

Connecting grants the agent access to Guru MCP Server on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Lago
Source: https://getskydive.com/docs/integrations/getlago

Open-source usage-based billing platform

## Example agent tasks

```text theme={null}
Connect Lago and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Lago.
```

```text theme={null}
Use the agent to interact with Lago directly from your conversation.
```

## Connecting

Connect Lago with an API key.

**1. Add the connection**

Generate an API key from your Lago account settings.

**2. Authorize**

Copy the key. It is only shown once.

**3. The agent gets to work**

Paste it when prompted in the agent's connection card for Lago.

## What you share

Connecting grants the agent access to Lago on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# GIPHY
Source: https://getskydive.com/docs/integrations/giphy

GIF search and sharing platform

## Example agent tasks

```text theme={null}
Connect GIPHY and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from GIPHY.
```

```text theme={null}
Use the agent to interact with GIPHY directly from your conversation.
```

## Connecting

Connect GIPHY with an API key.

**1. Add the connection**

Generate an API key from your GIPHY account settings.

**2. Authorize**

Copy the key. It is only shown once.

**3. The agent gets to work**

Paste it when prompted in the agent's connection card for GIPHY.

## What you share

Connecting grants the agent access to GIPHY on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# GitHub
Source: https://getskydive.com/docs/integrations/github

Give your agent its own GitHub identity

Agents can connect to GitHub with their **own bot identity**: so they can read and write code, open pull requests, and manage repositories, with their work attributed to them rather than to you.

## What the agent can do

> **Read & write code**
>
> Clone repositories, make changes, and push commits.

> **Open pull requests**
>
> Propose changes as PRs for your team to review.

> **Manage issues**
>
> Create, comment on, and update issues.

> **Run workflows**
>
> Work with Actions, checks, and statuses.

## Example agent tasks

```text theme={null}
When CI fails on main, pull the logs, find the likely cause, and open an issue with the details.
```

```text theme={null}
Every Monday, summarize the open pull requests and tell me which ones are waiting on me.
```

```text theme={null}
Review this pull request and comment on anything risky before I merge it.
```

## Its own identity

The connection gives the agent a dedicated bot identity on GitHub, with its **own name on commits and pull requests**. You can see at a glance what the agent did versus what a human did, and reviews stay clean. It's the agent's installation, scoped to the repositories you grant it.

## Connecting

**1. Authorize GitHub for the agent**

From the agent's connections, choose GitHub and install it on the org or the specific repositories you want it to access.

**2. Pick the repositories**

Grant access to all repos or only the ones the agent should touch.

**3. The agent gets to work**

It can now clone, commit, push, and open PRs against the repositories you granted, all under its own identity.

## If you're not an org owner

Connecting GitHub for an agent creates a dedicated GitHub App. When you connect it to an **organization**, GitHub only lets an owner (or an App Manager) create that app, so how you finish the install depends on your role.

### You're a GitHub App Manager

GitHub's membership API reports only owners, so if you hold the **App Manager** role rather than full ownership, the connect screen first shows your org as needing an owner. Choose **I'm an App Manager** and the org unlocks: you create the app yourself, and an org owner approves it afterward (the standard GitHub App approval, the same as any App Manager creating an app). Skydive remembers this, so next time your org is unlocked from the start.

### You're neither an owner nor an App Manager

Skydive hands off instead of hitting a dead end: copy the link from the connect screen and send it to an owner of your org. They open it, sign in to Skydive, and complete the GitHub install under their own identity. When they're done, the agent gets the same "GitHub connected" confirmation and can start working against the granted repositories.

**1. Start the connection**

Begin connecting GitHub for the agent and pick your organization. If GitHub doesn't already list you as an owner, you'll see both the App Manager option and a link to copy for an admin.

**2. Send the link to an org owner**

Copy it and pass it to someone who owns the org. The link explains which agent it's for and why access was requested.

**3. The owner completes the install**

They sign in to Skydive, continue to GitHub, and install the app on the org. The agent is notified and ready to go.

> [!TIP]
>
> Connecting GitHub for your **personal account** doesn't need a hand-off, you can do it yourself. The org paths above only apply to organization installs, which GitHub reserves for owners and App Managers.

## Scoped access

You control exactly which repositories the agent can reach and what it can do. Grant narrowly when you want a tight blast radius, or broadly for an agent that works across many repos. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# GitLab
Source: https://getskydive.com/docs/integrations/gitlab

Let your agent work with your GitLab projects

GitLab hosts your code, merge requests, and issues. Connect it and your agent can read and write across the projects you grant.

## What the agent can do

> **Read and write code**
>
> Work with files in a repository.

> **Merge requests**
>
> Open and review merge requests.

> **Issues**
>
> Create, comment on, and update issues.

> **Pipelines**
>
> Work with CI/CD pipelines and jobs.

## Example agent tasks

```text theme={null}
When a pipeline breaks, gather the logs and file an issue with what went wrong.
```

```text theme={null}
Each morning, tell me which merge requests are stuck waiting for review.
```

```text theme={null}
Summarize what changed across this week's merges for the release notes.
```

## What you share

During the connect flow you authorize what the agent can reach. It works only within what you grant. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Authorize GitLab for the agent**

From the agent's connections, choose GitLab and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[GitLab REST docs](https://docs.gitlab.com/ee/api/rest/)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Gitpod Classic Public
Source: https://getskydive.com/docs/integrations/gitpod

Cloud-based development environments

## Example agent tasks

```text theme={null}
Connect Gitpod Classic Public and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Gitpod Classic Public.
```

```text theme={null}
Use the agent to interact with Gitpod Classic Public directly from your conversation.
```

## Connecting

Connect Gitpod Classic Public with an API key.

**1. Add the connection**

Generate an API key from your Gitpod Classic Public account settings.

**2. Authorize**

Copy the key. It is only shown once.

**3. The agent gets to work**

Paste it when prompted in the agent's connection card for Gitpod Classic Public.

## What you share

Connecting grants the agent access to Gitpod Classic Public on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Glama MCP Inspector
Source: https://getskydive.com/docs/integrations/glama

MCP server registry and inspector

## Example agent tasks

```text theme={null}
Connect Glama MCP Inspector and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Glama MCP Inspector.
```

```text theme={null}
Use the agent to interact with Glama MCP Inspector directly from your conversation.
```

## Connecting

**MCP server**

Connect Glama Inspector over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Glama MCP Inspector.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Glama MCP Inspector through the MCP server.

**API key**

Connect Glama Gateway with an API key.

**1. Add the connection**

Generate an API key from your Glama MCP Inspector account settings.

**2. Authorize**

Copy the key. It is only shown once.

**3. The agent gets to work**

Paste it when prompted in the agent's connection card for Glama MCP Inspector.

## What you share

Connecting grants the agent access to Glama MCP Inspector on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Glean Developer Docs MCP Server
Source: https://getskydive.com/docs/integrations/glean

Enterprise search and work AI assistant

## Example agent tasks

```text theme={null}
Connect Glean Developer Docs MCP Server and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Glean Developer Docs MCP Server.
```

```text theme={null}
Use the agent to interact with Glean Developer Docs MCP Server directly from your conversation.
```

## Connecting

Connect Glean Developer Docs over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Glean Developer Docs.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Glean Developer Docs through the MCP server.

## What you share

Connecting grants the agent access to Glean Developer Docs MCP Server on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Gmail
Source: https://getskydive.com/docs/integrations/gmail

Let your agent read and send email from your Gmail

Gmail is your inbox. Connect it and your agent can read your mail, draft replies, and send messages on your behalf.

## What the agent can do

> **Read mail**
>
> Catch up on your inbox and pull a thread.

> **Draft and send**
>
> Prepare replies and send messages.

> **Search**
>
> Find an email by sender, subject, or content.

> **Organize**
>
> Read labels and sort your mail.

## Example agent tasks

```text theme={null}
Every morning, triage my inbox: summarize what matters, draft replies to the routine ones, and flag what's urgent.
```

```text theme={null}
When a contract arrives, pull the key dates into my calendar and file the email.
```

```text theme={null}
Watch for anything from this client and tell me the moment it lands.
```

## What you share

You authorize through Google's sign-in, and pick exactly which Google services to grant. Each is authorized on its own, so you share only what a task needs. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

> [!NOTE]
>
> Connecting this Google Workspace app (or any other approved Google Workspace app) locks this agent to [trusted models](/docs/agents/trusted-models) while the connection is active. Other models appear in the picker but cannot be selected. That is required under Google's Limited Use rules for user data.

## Connecting

**1. Authorize Gmail for the agent**

From the agent's connections, choose Gmail and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[Trusted models](/docs/agents/trusted-models)**
>
> Google Workspace connections limit which models this agent can use.

> **[Gmail API docs](https://developers.google.com/gmail/api)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Gocardless
Source: https://getskydive.com/docs/integrations/gocardless

Bank-to-bank direct debit payment collection

## Example agent tasks

```text theme={null}
Connect Gocardless and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Gocardless.
```

```text theme={null}
Use the agent to interact with Gocardless directly from your conversation.
```

## Connecting

Connect Gocardless over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Gocardless.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Gocardless through the MCP server.

## What you share

Connecting grants the agent access to Gocardless on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# GoDaddy MCP server
Source: https://getskydive.com/docs/integrations/godaddy

Fast, affordable domain and website setup for beginners

## Example agent tasks

```text theme={null}
Connect GoDaddy MCP server and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from GoDaddy MCP server.
```

```text theme={null}
Use the agent to interact with GoDaddy MCP server directly from your conversation.
```

## Connecting

Connect GoDaddy over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add GoDaddy.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach GoDaddy through the MCP server.

## What you share

Connecting grants the agent access to GoDaddy MCP server on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Jinko
Source: https://getskydive.com/docs/integrations/gojinko

Travel booking infrastructure for AI agents - search flights, hotels, and rental cars, then build and pay for trips

Jinko is a travel booking platform built for AI agents. Connect it and your agents can search live flights, hotels, and rental cars, assemble a trip, and hand off a checkout link for the user to pay. The full booking flow happens inside the conversation.

## What your agents can do

> **Search flights and hotels**
>
> Find live flight options, hotel inventory, and rental cars by route, date, and traveler count, with real pricing and availability.

> **Build and manage trips**
>
> Add flights, hotels, and ancillaries like baggage and seats to a shared trip cart. Adjust, remove, or reprice items before checkout.

> **Handle checkout and payment**
>
> Generate a checkout link the user opens to pay, or complete the booking server-side with a Shared Payment Token.

> **Manage bookings**
>
> Look up existing bookings, check refund eligibility, process refunds, and exchange booked flights for new itineraries.

## Example agent tasks

```text theme={null}
Search for round-trip flights from New York to Paris next month and show me the three cheapest options.
```

```text theme={null}
Add the selected flight and a hotel near the Eiffel Tower to my trip and send me a checkout link.
```

```text theme={null}
Check whether my flight on booking JNK-12345 is refundable and, if so, start the refund.
```

## What you share

Connecting gives your agents access to your Jinko account: the ability to search travel inventory, create and manage trips, and initiate bookings and refunds on your behalf. Payment requires your approval at checkout. Requires a Jinko account. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Add the Jinko connection**

From the agent's connections, choose Jinko. The platform connects through Jinko's MCP server at mcp.builders.gojinko.com.

**2. Sign in to Jinko**

A connect card appears. Sign in with your Jinko account at dashboard.gojinko.com and approve access.

**3. Your agents get to work**

They can now search flights, hotels, and cars, build trips, and manage bookings from inside any conversation.

## Learn more

> **[Jinko docs](https://docs.gojinko.com)**
>
> Full MCP tool reference, quickstart, and booking flow guides.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Gong
Source: https://getskydive.com/docs/integrations/gong

Let your agent work with your Gong call recordings and deal intelligence

Gong records, transcribes, and analyzes sales calls to help teams understand what drives deals. Connect it and your agent can surface call highlights, track deal health, and keep your CRM in sync with what's actually happening in conversations.

## What the agent can do

> **Pull call summaries**
>
> Retrieve transcripts, summaries, and key moments from recorded calls.

> **Track deals**
>
> Surface deal timelines, engagement patterns, and risk signals.

> **Research accounts**
>
> Pull call history and stakeholder activity for an account or opportunity.

> **Surface coaching insights**
>
> Identify talk patterns, objections, and themes across a rep's calls.

## Example agent tasks

```text theme={null}
Summarize the last three calls with Acme and tell me where the deal stands.
```

```text theme={null}
Which deals in my pipeline have had no customer engagement in the past two weeks?
```

```text theme={null}
Highlight the objections that came up on my calls this week.
```

## What you share

Connecting gives the agent access to your Gong workspace, including call recordings, transcripts, and deal data within your permissions. See [Permissions](/docs/integrations/permissions).

## Connecting

**1. Add the Gong connection**

From the agent's connections, choose Gong. The platform connects you through Gong's MCP server.

**2. Sign in to Gong**

A connect card appears. Sign in to your Gong workspace and approve access.

**3. The agent gets to work**

It can now pull call data, deal intelligence, and account activity from your Gong workspace.

## Learn more

> **[Gong help center](https://help.gong.io)**
>
> Product guides and API documentation.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Goodnotes Remote MCP Server
Source: https://getskydive.com/docs/integrations/goodnotes

Handwritten note-taking and PDF annotation for tablets

## Example agent tasks

```text theme={null}
Connect Goodnotes Remote MCP Server and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Goodnotes Remote MCP Server.
```

```text theme={null}
Use the agent to interact with Goodnotes Remote MCP Server directly from your conversation.
```

## Connecting

Connect Goodnotes Remote over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Goodnotes Remote.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Goodnotes Remote through the MCP server.

## What you share

Connecting grants the agent access to Goodnotes Remote MCP Server on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Google
Source: https://getskydive.com/docs/integrations/google

Google account identity and core services

## Example agent tasks

```text theme={null}
Connect Google and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Google.
```

```text theme={null}
Use the agent to interact with Google directly from your conversation.
```

## Connecting

Connect your Google account so your agent can work in it.

**1. Add the connection**

From the agent's connections, choose Google and sign in.

**2. Authorize**

Approve the access. You can change this later.

**3. The agent gets to work**

The agent gets to work through the secure proxy.

## What you share

Connecting grants the agent access to Google on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Ads
Source: https://getskydive.com/docs/integrations/google-ads

Let your agent read your Google Ads performance

Google Ads runs your search and display campaigns. Connect it and your agent can read your campaigns and their performance.

## What the agent can do

> **Read campaigns**
>
> See your campaigns and ad groups.

> **See performance**
>
> Pull clicks, spend, and conversions.

> **Find data**
>
> Look up specific campaigns.

> **See spend**
>
> Read budget and cost.

## Example agent tasks

```text theme={null}
Every Monday, tell me how my campaigns did and where to shift budget.
```

```text theme={null}
When a campaign's cost spikes, dig into why and alert me.
```

```text theme={null}
Tell me which campaign converts best and what to scale.
```

## What you share

You authorize through Google's sign-in, and pick exactly which Google services to grant. Each is authorized on its own, so you share only what a task needs. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Authorize Ads for the agent**

From the agent's connections, choose Ads and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[Google Ads API docs](https://developers.google.com/google-ads/api/docs/start)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Analytics
Source: https://getskydive.com/docs/integrations/google-analytics

Let your agent read your Google Analytics data

Google Analytics tracks your website traffic. Connect it and your agent can read your traffic, sources, and trends.

## What the agent can do

> **Read metrics**
>
> Pull traffic, users, and sessions.

> **See sources**
>
> Read where visitors come from.

> **By date**
>
> Read data over a time range.

> **Find reports**
>
> Look up specific reports.

## Example agent tasks

```text theme={null}
Every Monday, send me the traffic story: what moved, what's growing, what dropped.
```

```text theme={null}
When traffic spikes, tell me where it came from so I can lean in.
```

```text theme={null}
Compare this month to last and explain what changed.
```

## What you share

You authorize through Google's sign-in, and pick exactly which Google services to grant. Each is authorized on its own, so you share only what a task needs. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Authorize Analytics for the agent**

From the agent's connections, choose Analytics and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[Google Analytics API docs](https://developers.google.com/analytics)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Calendar
Source: https://getskydive.com/docs/integrations/google-calendar

Let your agent read and manage your Google Calendar

Google Calendar holds your schedule. Connect it and your agent can read your events and create new ones.

## What the agent can do

> **Read events**
>
> See what's on your calendar.

> **Create events**
>
> Add meetings and reminders.

> **Check availability**
>
> See when you're free or busy.

> **Find events**
>
> Search your schedule.

## Example agent tasks

```text theme={null}
Each morning, brief me on my day with context on everyone I'm meeting.
```

```text theme={null}
When someone asks to meet, find a slot that works and send the invite.
```

```text theme={null}
Tell me where my week is overbooked and what I could reschedule.
```

## What you share

You authorize through Google's sign-in, and pick exactly which Google services to grant. Each is authorized on its own, so you share only what a task needs. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

> [!NOTE]
>
> Connecting this Google Workspace app (or any other approved Google Workspace app) locks this agent to [trusted models](/docs/agents/trusted-models) while the connection is active. Other models appear in the picker but cannot be selected. That is required under Google's Limited Use rules for user data.

## Connecting

**1. Authorize Calendar for the agent**

From the agent's connections, choose Calendar and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[Trusted models](/docs/agents/trusted-models)**
>
> Google Workspace connections limit which models this agent can use.

> **[Google Calendar API docs](https://developers.google.com/calendar/api)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Contacts
Source: https://getskydive.com/docs/integrations/google-contacts

Let your agent read your Google Contacts

Google Contacts holds your address book. Connect it and your agent can look up people and their details.

## What the agent can do

> **Read contacts**
>
> Look up a person's details.

> **Search**
>
> Find a contact by name or company.

> **Browse**
>
> See your contact list.

> **Get details**
>
> Pull email and phone.

## Example agent tasks

```text theme={null}
Before a meeting, pull up who I'm meeting and our recent history.
```

```text theme={null}
Find everyone I know at this company and my warmest intro.
```

```text theme={null}
Keep this client's details current when they change.
```

## What you share

You authorize through Google's sign-in, and pick exactly which Google services to grant. Each is authorized on its own, so you share only what a task needs. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

> [!NOTE]
>
> Connecting this Google Workspace app (or any other approved Google Workspace app) locks this agent to [trusted models](/docs/agents/trusted-models) while the connection is active. Other models appear in the picker but cannot be selected. That is required under Google's Limited Use rules for user data.

## Connecting

**1. Authorize Contacts for the agent**

From the agent's connections, choose Contacts and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[Trusted models](/docs/agents/trusted-models)**
>
> Google Workspace connections limit which models this agent can use.

> **[Google People API docs](https://developers.google.com/people)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Docs
Source: https://getskydive.com/docs/integrations/google-docs

Let your agent read and write Google Docs

Google Docs is your word processor. Connect it and your agent can read documents and create or update them.

## What the agent can do

> **Read docs**
>
> Pull the content of a document.

> **Create docs**
>
> Write a new document.

> **Update docs**
>
> Edit the contents of a doc.

> **Find a doc**
>
> Locate the right document.

## Example agent tasks

```text theme={null}
Draft the proposal from this brief, matching the tone of our last three.
```

```text theme={null}
Turn my meeting notes into a clean recap and share it with the team.
```

```text theme={null}
Pull the action items out of this doc and add them to my task list.
```

## What you share

You authorize through Google's sign-in, and pick exactly which Google services to grant. Each is authorized on its own, so you share only what a task needs. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

> [!NOTE]
>
> Connecting this Google Workspace app (or any other approved Google Workspace app) locks this agent to [trusted models](/docs/agents/trusted-models) while the connection is active. Other models appear in the picker but cannot be selected. That is required under Google's Limited Use rules for user data.

## Connecting

**1. Authorize Docs for the agent**

From the agent's connections, choose Docs and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[Trusted models](/docs/agents/trusted-models)**
>
> Google Workspace connections limit which models this agent can use.

> **[Google Docs API docs](https://developers.google.com/docs/api)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Drive
Source: https://getskydive.com/docs/integrations/google-drive

Let your agent work with your Google Drive files

Google Drive stores your files. Connect it and your agent can find, read, and save documents.

## What the agent can do

> **Find files**
>
> Search your Drive by name or folder.

> **Read files**
>
> Pull content from a document.

> **Save files**
>
> Write new files to a folder.

> **Browse folders**
>
> Move through your Drive.

## Example agent tasks

```text theme={null}
When a signed doc lands, file it in the right folder and log the renewal date.
```

```text theme={null}
Every Friday, gather the week's work into a single recap doc.
```

```text theme={null}
Find every file mentioning the old pricing and list what needs updating.
```

## What you share

You authorize through Google's sign-in, and pick exactly which Google services to grant. Each is authorized on its own, so you share only what a task needs. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

> [!NOTE]
>
> Connecting this Google Workspace app (or any other approved Google Workspace app) locks this agent to [trusted models](/docs/agents/trusted-models) while the connection is active. Other models appear in the picker but cannot be selected. That is required under Google's Limited Use rules for user data.

## Connecting

**1. Authorize Drive for the agent**

From the agent's connections, choose Drive and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[Trusted models](/docs/agents/trusted-models)**
>
> Google Workspace connections limit which models this agent can use.

> **[Google Drive API docs](https://developers.google.com/drive/api)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Forms
Source: https://getskydive.com/docs/integrations/google-forms

Let your agent read your Google Forms and responses

Google Forms collects survey and form responses. Connect it and your agent can read your forms and the answers people submit.

## What the agent can do

> **Read forms**
>
> See your form structure.

> **Read responses**
>
> Pull submitted answers.

> **Summarize**
>
> Roll up the results.

> **Find a form**
>
> Locate the right form.

## Example agent tasks

```text theme={null}
When someone submits the lead form, qualify it and route a hot one to me now.
```

```text theme={null}
Every Friday, summarize the week's responses and the themes to act on.
```

```text theme={null}
Flag any answer that signals an unhappy customer.
```

## What you share

You authorize through Google's sign-in, and pick exactly which Google services to grant. Each is authorized on its own, so you share only what a task needs. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

> [!NOTE]
>
> Connecting this Google Workspace app (or any other approved Google Workspace app) locks this agent to [trusted models](/docs/agents/trusted-models) while the connection is active. Other models appear in the picker but cannot be selected. That is required under Google's Limited Use rules for user data.

## Connecting

**1. Authorize Forms for the agent**

From the agent's connections, choose Forms and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[Trusted models](/docs/agents/trusted-models)**
>
> Google Workspace connections limit which models this agent can use.

> **[Google Forms API docs](https://developers.google.com/forms/api)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Google Mail
Source: https://getskydive.com/docs/integrations/google-mail

Email inbox and messaging

## Example agent tasks

```text theme={null}
Connect Google Mail and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Google Mail.
```

```text theme={null}
Use the agent to interact with Google Mail directly from your conversation.
```

## Connecting

Connect your Google Mail account so your agent can work in it.

**1. Add the connection**

From the agent's connections, choose Google Mail and sign in.

**2. Authorize**

Approve the access. You can change this later.

**3. The agent gets to work**

The agent gets to work through the secure proxy.

## What you share

Connecting grants the agent access to Google Mail on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Meet
Source: https://getskydive.com/docs/integrations/google-meet

Let your agent work with your Google Meet meetings

Google Meet runs your video calls. Connect it and your agent can read meeting details and recordings.

## What the agent can do

> **Read meetings**
>
> See your meeting details.

> **Recordings**
>
> Pull recordings and transcripts where available.

> **See schedule**
>
> Read upcoming meetings.

> **Participants**
>
> See who attended.

## Example agent tasks

```text theme={null}
After each call, send me the summary, decisions, and action items.
```

```text theme={null}
When I miss a meeting, catch me up from the recording in two minutes.
```

```text theme={null}
Every Friday, pull the week's call notes into one recap.
```

## What you share

You authorize through Google's sign-in, and pick exactly which Google services to grant. Each is authorized on its own, so you share only what a task needs. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Authorize Meet for the agent**

From the agent's connections, choose Meet and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[Google Meet API docs](https://developers.google.com/meet/api)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Play
Source: https://getskydive.com/docs/integrations/google-play

Let your agent work with your Google Play data

Google Play is the Android app store. Connect it and your agent can read your app's listing and release data.

## What the agent can do

> **Read app data**
>
> See your app's listing and details.

> **See releases**
>
> Read release and rollout status.

> **See reviews**
>
> Read user reviews and ratings.

> **See stats**
>
> Roll up install and rating data.

## Example agent tasks

```text theme={null}
When a release rolls out, watch the reviews and flag any spike in complaints.
```

```text theme={null}
Every Monday, summarize my rating, installs, and what reviewers are saying.
```

```text theme={null}
Tell me the moment a review mentions a crash or a bug.
```

## What you share

You authorize through Google's sign-in, and pick exactly which Google services to grant. Each is authorized on its own, so you share only what a task needs. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Authorize Play for the agent**

From the agent's connections, choose Play and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[Google Play Developer API docs](https://developers.google.com/android-publisher)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Search Console
Source: https://getskydive.com/docs/integrations/google-search-console

Let your agent read your Search Console data

Google Search Console reports how your site does in search. Connect it and your agent can read your search performance and indexing.

## What the agent can do

> **Read performance**
>
> Pull clicks, impressions, and rankings.

> **See queries**
>
> Read which searches bring traffic.

> **Check indexing**
>
> See how pages are indexed.

> **By date**
>
> Read data over a range.

## Example agent tasks

```text theme={null}
Every Monday, tell me which search queries are gaining and which pages are slipping.
```

```text theme={null}
When a page falls in rankings, flag it and suggest why.
```

```text theme={null}
Find the searches we rank for but don't have content on.
```

## What you share

You authorize through Google's sign-in, and pick exactly which Google services to grant. Each is authorized on its own, so you share only what a task needs. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Authorize Search Console for the agent**

From the agent's connections, choose Search Console and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[Search Console API docs](https://developers.google.com/webmaster-tools)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Google Sheet
Source: https://getskydive.com/docs/integrations/google-sheet

Online spreadsheets

## Example agent tasks

```text theme={null}
Connect Google Sheet and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Google Sheet.
```

```text theme={null}
Use the agent to interact with Google Sheet directly from your conversation.
```

## Connecting

Connect your Google Sheet account so your agent can work in it.

**1. Add the connection**

From the agent's connections, choose Google Sheet and sign in.

**2. Authorize**

Approve the access. You can change this later.

**3. The agent gets to work**

The agent gets to work through the secure proxy.

## What you share

Connecting grants the agent access to Google Sheet on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Sheets
Source: https://getskydive.com/docs/integrations/google-sheets

Let your agent read and update your Google Sheets

Google Sheets is your spreadsheet. Connect it and your agent can read cells and ranges and write values back.

## What the agent can do

> **Read data**
>
> Pull values from a sheet or range.

> **Write values**
>
> Update cells and add rows.

> **Work with ranges**
>
> Read and filter structured data.

> **Find a sheet**
>
> Locate the right spreadsheet.

## Example agent tasks

```text theme={null}
Every Monday, refresh my tracker from the latest data and flag what's slipping.
```

```text theme={null}
When a lead comes in, add it to my sheet with the source and a follow-up date.
```

```text theme={null}
Turn this raw export into a clean summary with totals by category.
```

## What you share

You authorize through Google's sign-in, and pick exactly which Google services to grant. Each is authorized on its own, so you share only what a task needs. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

> [!NOTE]
>
> Connecting this Google Workspace app (or any other approved Google Workspace app) locks this agent to [trusted models](/docs/agents/trusted-models) while the connection is active. Other models appear in the picker but cannot be selected. That is required under Google's Limited Use rules for user data.

## Connecting

**1. Authorize Sheets for the agent**

From the agent's connections, choose Sheets and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[Trusted models](/docs/agents/trusted-models)**
>
> Google Workspace connections limit which models this agent can use.

> **[Google Sheets API docs](https://developers.google.com/sheets/api)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Slides
Source: https://getskydive.com/docs/integrations/google-slides

Let your agent work with your Google Slides

Google Slides is your presentation tool. Connect it and your agent can read your decks and create new ones.

## What the agent can do

> **Read decks**
>
> Pull content from a presentation.

> **Create decks**
>
> Build a new presentation.

> **Update slides**
>
> Edit slides in a deck.

> **Find a deck**
>
> Locate the right presentation.

## Example agent tasks

```text theme={null}
Build a first-draft deck from this quarter's numbers and our usual template.
```

```text theme={null}
Refresh the metrics slides in my monthly deck with the latest data.
```

```text theme={null}
Pull the talking points so I can rehearse from my phone.
```

## What you share

You authorize through Google's sign-in, and pick exactly which Google services to grant. Each is authorized on its own, so you share only what a task needs. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

> [!NOTE]
>
> Connecting this Google Workspace app (or any other approved Google Workspace app) locks this agent to [trusted models](/docs/agents/trusted-models) while the connection is active. Other models appear in the picker but cannot be selected. That is required under Google's Limited Use rules for user data.

## Connecting

**1. Authorize Slides for the agent**

From the agent's connections, choose Slides and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[Trusted models](/docs/agents/trusted-models)**
>
> Google Workspace connections limit which models this agent can use.

> **[Google Slides API docs](https://developers.google.com/slides/api)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Tasks
Source: https://getskydive.com/docs/integrations/google-tasks

Let your agent manage your Google Tasks

Google Tasks is your to-do list in Google Workspace. Connect it and your agent can read tasks and create or complete them.

## What the agent can do

> **Read tasks**
>
> See your tasks and lists.

> **Create tasks**
>
> Add tasks with due dates.

> **Complete tasks**
>
> Mark tasks done.

> **See lists**
>
> Find the right task list.

## Example agent tasks

```text theme={null}
Each morning, plan my day from what's due and what I keep pushing.
```

```text theme={null}
Turn the action items from my last meeting into tasks with deadlines.
```

```text theme={null}
At week's end, tell me what I finished and what slipped.
```

## What you share

You authorize through Google's sign-in, and pick exactly which Google services to grant. Each is authorized on its own, so you share only what a task needs. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

> [!NOTE]
>
> Connecting this Google Workspace app (or any other approved Google Workspace app) locks this agent to [trusted models](/docs/agents/trusted-models) while the connection is active. Other models appear in the picker but cannot be selected. That is required under Google's Limited Use rules for user data.

## Connecting

**1. Authorize Tasks for the agent**

From the agent's connections, choose Tasks and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[Trusted models](/docs/agents/trusted-models)**
>
> Google Workspace connections limit which models this agent can use.

> **[Google Tasks API docs](https://developers.google.com/tasks)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Gorgias
Source: https://getskydive.com/docs/integrations/gorgias

Customer support helpdesk for ecommerce

## Example agent tasks

```text theme={null}
Connect Gorgias and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Gorgias.
```

```text theme={null}
Use the agent to interact with Gorgias directly from your conversation.
```

## Connecting

Connect Gorgias over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Gorgias.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Gorgias through the MCP server.

## What you share

Connecting grants the agent access to Gorgias on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Grafana
Source: https://getskydive.com/docs/integrations/grafana

Dashboards and observability for metrics

Grafana is the standard platform for dashboards and observability. Connect Grafana Cloud and your agent can query metrics, inspect dashboards, and surface what's happening across your systems without you opening a browser tab.

## What the agent can do

> **Query metrics and data**
>
> Run PromQL and other queries against your Grafana Cloud data sources to pull live metrics.

> **Inspect dashboards**
>
> Read panel definitions, current values, and alerts configured in your dashboards.

> **Check alert status**
>
> See which alerts are firing, their severity, and the conditions that triggered them.

> **Explore logs and traces**
>
> Query Loki logs and Tempo traces to investigate issues across your stack.

## Example agent tasks

```text theme={null}
What's the current p99 latency for the payments service over the last hour?
```

```text theme={null}
Are there any firing alerts right now? Summarize the most critical ones.
```

```text theme={null}
Show me error rate trends for the API gateway over the past 24 hours.
```

## What you share

Connecting gives the agent access to your Grafana Cloud workspace, including dashboards, data sources, metrics, logs, and alert configurations. See [Permissions](/docs/integrations/permissions).

## Connecting

**1. Add the Grafana connection**

From the agent's connections, choose Grafana. The platform connects you through Grafana Cloud's MCP server.

**2. Sign in to Grafana Cloud**

A connect card appears. Sign in to your Grafana Cloud account and approve access.

**3. The agent gets to work**

It can now query metrics, read dashboards, and surface alert status from your Grafana Cloud workspace.

## Learn more

> **[Grafana documentation](https://grafana.com/docs)**
>
> Grafana Cloud documentation and guides.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Grain
Source: https://getskydive.com/docs/integrations/grain

Let your agent work with your Grain meeting recordings and highlights

Grain records meetings, clips highlights, and makes call content shareable. Connect it and your agent can surface transcripts, pull highlight reels, and search across your recorded conversations.

## What the agent can do

> **Pull meeting transcripts**
>
> Retrieve full transcripts from recorded meetings in your Grain workspace.

> **Surface highlights**
>
> Find and share highlight clips from specific moments in a call.

> **Search recordings**
>
> Search across your Grain library by topic, speaker, or keyword.

> **Share content**
>
> Surface meeting clips and summaries to share with teammates.

## Example agent tasks

```text theme={null}
Pull the highlights from this week's customer discovery calls and summarize the common themes.
```

```text theme={null}
Find the clip from the Acme call where they talked about their timeline.
```

```text theme={null}
What did the CEO say in the all-hands last week? Pull the key points.
```

## What you share

Connecting gives the agent access to your Grain workspace, including recordings, transcripts, and highlights within your account. See [Permissions](/docs/integrations/permissions).

## Connecting

**1. Add the Grain connection**

From the agent's connections, choose Grain. The platform connects you through Grain's MCP server.

**2. Sign in to Grain**

A connect card appears. Sign in to your Grain account and approve access.

**3. The agent gets to work**

It can now pull transcripts, highlights, and recordings from your Grain workspace.

## Learn more

> **[Grain help center](https://grain.com/help)**
>
> Product guides and how-tos.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Granola
Source: https://getskydive.com/docs/integrations/granola

AI notepad for meetings

## Example agent tasks

```text theme={null}
Connect Granola and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Granola.
```

```text theme={null}
Use the agent to interact with Granola directly from your conversation.
```

## Connecting

Connect Granola over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Granola.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Granola through the MCP server.

## What you share

Connecting grants the agent access to Granola on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Gremlin
Source: https://getskydive.com/docs/integrations/gremlin

Chaos engineering and resilience testing

## Example agent tasks

```text theme={null}
Connect Gremlin and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Gremlin.
```

```text theme={null}
Use the agent to interact with Gremlin directly from your conversation.
```

## Connecting

Connect Gremlin with an API key.

**1. Add the connection**

Generate an API key from your Gremlin account settings.

**2. Authorize**

Copy the key. It is only shown once.

**3. The agent gets to work**

Paste it when prompted in the agent's connection card for Gremlin.

## What you share

Connecting grants the agent access to Gremlin on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Gumloop
Source: https://getskydive.com/docs/integrations/gumloop

Let your agent build and run Gumloop AI workflows

Gumloop is a no-code platform for building AI-powered automation workflows. Connect it and your agent can trigger runs, pass inputs, and retrieve results from your Gumloop automations.

## What the agent can do

> **Trigger workflow runs**
>
> Start a Gumloop workflow and pass custom inputs into it.

> **Retrieve run outputs**
>
> Pull results and structured output from completed workflow runs.

> **Inspect flow status**
>
> Check whether a run completed, errored, or is still in progress.

> **Chain automations**
>
> Use Gumloop as a step in a broader agent workflow, passing data in and out.

## Example agent tasks

```text theme={null}
Run my lead enrichment flow with this company name and return the results.
```

```text theme={null}
Trigger the daily newsletter pipeline and confirm it finished without errors.
```

```text theme={null}
Start the document summarization workflow on this URL and tell me what it found.
```

## What you share

Connecting gives the agent access to your Gumloop workspace, including your flows and their run history. See [Permissions](/docs/integrations/permissions).

## Connecting

**1. Add the Gumloop connection**

From the agent's connections, choose Gumloop. The platform connects you through Gumloop's hosted MCP server.

**2. Sign in to Gumloop**

A connect card appears. Sign in to your Gumloop account and approve access.

**3. The agent gets to work**

It can now trigger and retrieve results from your Gumloop workflows.

## Learn more

> **[Gumloop documentation](https://docs.gumloop.com)**
>
> Platform guides and workflow reference.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Gusto
Source: https://getskydive.com/docs/integrations/gusto

Payroll for lean small teams who want it simple

## Example agent tasks

```text theme={null}
Connect Gusto and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Gusto.
```

```text theme={null}
Use the agent to interact with Gusto directly from your conversation.
```

## Connecting

Connect Gusto over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Gusto.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Gusto through the MCP server.

## What you share

Connecting grants the agent access to Gusto on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Habitify
Source: https://getskydive.com/docs/integrations/habitify

Habit tracking and daily routines

## Example agent tasks

```text theme={null}
Connect Habitify and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Habitify.
```

```text theme={null}
Use the agent to interact with Habitify directly from your conversation.
```

## Connecting

Connect Habitify over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Habitify.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Habitify through the MCP server.

## What you share

Connecting grants the agent access to Habitify on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Harvest
Source: https://getskydive.com/docs/integrations/harvest

Let your agent work with your Harvest time tracking

Harvest tracks time and invoices for client work. Connect it and your agent can read time entries, projects, and invoices.

## What the agent can do

> **Read time**
>
> See hours logged by project or person.

> **See projects**
>
> Read your projects and clients.

> **Invoices**
>
> Read invoices and their status.

> **Summaries**
>
> Roll up tracked time.

## Example agent tasks

```text theme={null}
How many hours did we log on the Acme project this week?
```

```text theme={null}
Every Friday, tell me what's billable and what's unbilled.
```

```text theme={null}
Which projects are over their estimated hours?
```

## What you share

During the connect flow you authorize what the agent can reach. It works only within what you grant. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Authorize Harvest for the agent**

From the agent's connections, choose Harvest and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[Harvest API docs](https://help.getharvest.com/api-v2/)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Hatchet Cloud
Source: https://getskydive.com/docs/integrations/hatchet

Background task queue and workflow orchestration

## Example agent tasks

```text theme={null}
Connect Hatchet Cloud and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Hatchet Cloud.
```

```text theme={null}
Use the agent to interact with Hatchet Cloud directly from your conversation.
```

## Connecting

Connect Hatchet Cloud with an API key.

**1. Add the connection**

Generate an API key from your Hatchet Cloud account settings.

**2. Authorize**

Copy the key. It is only shown once.

**3. The agent gets to work**

Paste it when prompted in the agent's connection card for Hatchet Cloud.

## What you share

Connecting grants the agent access to Hatchet Cloud on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Help Scout
Source: https://getskydive.com/docs/integrations/helpscout

Let your agent work with your Help Scout conversations

Help Scout is a help desk built around a shared inbox. Connect it and your agent can read conversations and help reply.

## What the agent can do

> **Read conversations**
>
> Catch up on support threads.

> **Read customers**
>
> Pull a customer's details.

> **Draft replies**
>
> Prepare responses for your team.

> **Search**
>
> Find the right conversation.

## Example agent tasks

```text theme={null}
Every morning, summarize new conversations and draft answers to the common ones.
```

```text theme={null}
When a customer's been waiting too long, flag it before they churn.
```

```text theme={null}
Pull a customer's full history before I reply so I have the context.
```

## What you share

During the connect flow you authorize what the agent can reach. It works only within what you grant. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Authorize Help Scout for the agent**

From the agent's connections, choose Help Scout and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[Help Scout Mailbox API docs](https://developer.helpscout.com/mailbox-api/)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Help Scout Mailbox
Source: https://getskydive.com/docs/integrations/helpscout-mailbox

Shared inbox for customer support

## Example agent tasks

```text theme={null}
Connect Help Scout Mailbox and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Help Scout Mailbox.
```

```text theme={null}
Use the agent to interact with Help Scout Mailbox directly from your conversation.
```

## Connecting

Connect your Help Scout Mailbox account so your agent can work in it.

**1. Add the connection**

From the agent's connections, choose Help Scout Mailbox and sign in.

**2. Authorize**

Approve the access. You can change this later.

**3. The agent gets to work**

The agent gets to work through the secure proxy.

## What you share

Connecting grants the agent access to Help Scout Mailbox on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Hercules
Source: https://getskydive.com/docs/integrations/hercules

AI app builder for internal business software

## Example agent tasks

```text theme={null}
Connect Hercules and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Hercules.
```

```text theme={null}
Use the agent to interact with Hercules directly from your conversation.
```

## Connecting

Connect Hercules over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Hercules.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Hercules through the MCP server.

## What you share

Connecting grants the agent access to Hercules on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Hex
Source: https://getskydive.com/docs/integrations/hex

Collaborative data notebooks and analytics

## Example agent tasks

```text theme={null}
Connect Hex and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Hex.
```

```text theme={null}
Use the agent to interact with Hex directly from your conversation.
```

## Connecting

Connect Hex over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Hex.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Hex through the MCP server.

## What you share

Connecting grants the agent access to Hex on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# hotglue
Source: https://getskydive.com/docs/integrations/hotglue

Embedded data integration for SaaS apps

## Example agent tasks

```text theme={null}
Connect hotglue and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from hotglue.
```

```text theme={null}
Use the agent to interact with hotglue directly from your conversation.
```

## Connecting

Connect hotglue with an API key.

**1. Add the connection**

Generate an API key from your hotglue account settings.

**2. Authorize**

Copy the key. It is only shown once.

**3. The agent gets to work**

Paste it when prompted in the agent's connection card for hotglue.

## What you share

Connecting grants the agent access to hotglue on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# HubSpot REST APIs
Source: https://getskydive.com/docs/integrations/hubapi

CRM and marketing automation platform

## Example agent tasks

```text theme={null}
Connect HubSpot REST APIs and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from HubSpot REST APIs.
```

```text theme={null}
Use the agent to interact with HubSpot REST APIs directly from your conversation.
```

## Connecting

Connect HubSpot REST APIs with an API key.

**1. Add the connection**

Generate an API key from your HubSpot REST APIs account settings.

**2. Authorize**

Copy the key. It is only shown once.

**3. The agent gets to work**

Paste it when prompted in the agent's connection card for HubSpot REST APIs.

## What you share

Connecting grants the agent access to HubSpot REST APIs on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# HubSpot
Source: https://getskydive.com/docs/integrations/hubspot

Let your agents manage contacts, deals, and records in HubSpot

HubSpot is a CRM platform for managing contacts, companies, deals, and marketing. Connect it and your agents can read and update CRM records, track deals, and work across your pipeline.

## Two ways to connect

HubSpot supports a private-app token (recommended) and OAuth sign-in.

* **HubSpot API (recommended).** A private-app token from your HubSpot account. No OAuth consent flow required, and you control exactly which scopes the token carries. See [Connect with HubSpot API](#connect-with-hubspot-api).
* **HubSpot OAuth.** Sign in with your HubSpot account. See [Connect with OAuth](#connect-with-oauth).

## What the agent can do

> **Read contacts**
>
> Pull contact records, properties, and activity history.

> **Manage deals**
>
> Create, update, and move deals through your pipeline stages.

> **Update records**
>
> Edit companies, contacts, and other CRM objects you grant access to.

> **Search the CRM**
>
> Find records by property, owner, stage, or custom filters.

## Example agent tasks

```text theme={null}
Show me all open deals in the proposal stage and who owns each one.
```

```text theme={null}
Every Monday morning, list contacts added last week that have no follow-up scheduled.
```

```text theme={null}
Update this deal's stage to Closed Won and log a note with the agreed amount.
```

## Connect with HubSpot API

Create a private app in HubSpot, grant it the scopes your agents need, and paste the access token.

**1. Create a private app**

Go to [https://app.hubspot.com/private-apps](https://app.hubspot.com/private-apps), create a new private app, and grant the CRM scopes you want the agent to have.

**2. Copy the access token**

After saving, copy the access token from the Auth tab. It is only shown once.

**3. Add the connection**

From the agent's connections, choose HubSpot and paste the token when prompted.

**4. The agent gets to work**

It calls the HubSpot CRM REST API directly, within the scopes the private app was granted.

## Connect with OAuth

## What you share

During the connect flow you authorize what the agent can reach. It works only within what you grant. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

**1. Authorize HubSpot for the agent**

From the agent's connections, choose HubSpot and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[HubSpot API docs](https://developers.hubspot.com/docs/reference/api/overview)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Huggingface
Source: https://getskydive.com/docs/integrations/huggingface

Open-source AI model and dataset hub

## Example agent tasks

```text theme={null}
Connect Huggingface and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Huggingface.
```

```text theme={null}
Use the agent to interact with Huggingface directly from your conversation.
```

## Connecting

Connect Huggingface over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Huggingface.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Huggingface through the MCP server.

## What you share

Connecting grants the agent access to Huggingface on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# incident.io
Source: https://getskydive.com/docs/integrations/incident

Incident management, on-call, and AI investigations, connected over MCP

incident.io runs incident response, on-call schedules, and alert routing, and Investigations adds AI analysis over your incidents, alerts, and telemetry. Connect it over MCP and your agents can manage incidents and escalations, check schedules, query logs and metrics, and pull investigation findings, with about 90 tools.

## What your agents can do

> **Manage incidents**
>
> Create and update incidents, merge duplicates, post findings to the incident channel, and track actions and follow-ups.

> **Work on-call and escalations**
>
> Check schedules and who is on call, acknowledge or page, set overrides, and hold back paging with maintenance windows.

> **Analyze incidents and alerts**
>
> incident\_stats, alert\_stats, and duration\_metrics\_stats group counts and responder workload by severity, team, service, and time period.

> **Run investigations and telemetry queries**
>
> Start or steer an AI investigation, download its findings as an archive, and query logs, metrics, and traces across Datadog, Grafana, and other connected platforms.

## Example agent tasks

```text theme={null}
Check who is on call for the platform schedule, then summarize every open SEV-1 by workload.
```

```text theme={null}
Show me the top alert sources behind our overnight pages and which ones turned into incidents.
```

```text theme={null}
Download the investigation for INC-123 and pull the error-rate logs around the failure window.
```

## Connecting

**Connect Incident.io**

Sign in with your incident.io account through the OAuth flow on the connect card. No API key or developer setup needed. You need an incident.io account.

**1. Add the incident.io connection**

From your agent's connections, choose incident.io.

**2. Sign in to incident.io**

A connect card appears. Sign in to your incident.io account and approve access.

**3. The agent gets to work**

It can now manage incidents, work schedules, and run analysis across your incident data.

The connection acts as the person who approves it, so agents see and do what that account can. The remote MCP server is available on incident.io's Team, Pro, and Enterprise plans; it is not on the Basic (free) plan.

## What you share

Connecting grants your agents access to incident.io on your behalf, scoped to your account. They can read and write incidents, alerts, schedules, escalations, and status page drafts. Status page updates stay drafts; publishing one emails and texts every subscriber, so a person always does that part. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

## Learn more

> **[incident.io remote MCP documentation](https://docs.incident.io/ai/remote-mcp)**
>
> The full tool list and setup guide from incident.io.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Increase
Source: https://getskydive.com/docs/integrations/increase

Banking infrastructure and payments API

## Example agent tasks

```text theme={null}
Connect Increase and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Increase.
```

```text theme={null}
Use the agent to interact with Increase directly from your conversation.
```

## Connecting

Connect Increase with an API key.

**1. Add the connection**

Generate an API key from your Increase account settings.

**2. Authorize**

Copy the key. It is only shown once.

**3. The agent gets to work**

Paste it when prompted in the agent's connection card for Increase.

## What you share

Connecting grants the agent access to Increase on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# IP2Proxy Proxy Detection
Source: https://getskydive.com/docs/integrations/ip2proxy

Proxy and VPN detection API

## Example agent tasks

```text theme={null}
Connect IP2Proxy Proxy Detection and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from IP2Proxy Proxy Detection.
```

```text theme={null}
Use the agent to interact with IP2Proxy Proxy Detection directly from your conversation.
```

## Connecting

Connect IP2Proxy Proxy Detection with an API key.

**1. Add the connection**

Generate an API key from your IP2Proxy Proxy Detection account settings.

**2. Authorize**

Copy the key. It is only shown once.

**3. The agent gets to work**

Paste it when prompted in the agent's connection card for IP2Proxy Proxy Detection.

## What you share

Connecting grants the agent access to IP2Proxy Proxy Detection on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# iubenda Consent Database
Source: https://getskydive.com/docs/integrations/iubenda

Privacy policy and cookie consent management

## Example agent tasks

```text theme={null}
Connect iubenda Consent Database and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from iubenda Consent Database.
```

```text theme={null}
Use the agent to interact with iubenda Consent Database directly from your conversation.
```

## Connecting

Connect iubenda Consent Database with an API key.

**1. Add the connection**

Generate an API key from your iubenda Consent Database account settings.

**2. Authorize**

Copy the key. It is only shown once.

**3. The agent gets to work**

Paste it when prompted in the agent's connection card for iubenda Consent Database.

## What you share

Connecting grants the agent access to iubenda Consent Database on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Jam
Source: https://getskydive.com/docs/integrations/jam

Bug reporting for developers

## Example agent tasks

```text theme={null}
Connect Jam and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Jam.
```

```text theme={null}
Use the agent to interact with Jam directly from your conversation.
```

## Connecting

Connect Jam over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Jam.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Jam through the MCP server.

## What you share

Connecting grants the agent access to Jam on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Jimdo
Source: https://getskydive.com/docs/integrations/jimdo

AI website builder for solopreneurs, made in Germany

## Example agent tasks

```text theme={null}
Connect Jimdo and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Jimdo.
```

```text theme={null}
Use the agent to interact with Jimdo directly from your conversation.
```

## Connecting

Connect Jimdo over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Jimdo.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Jimdo through the MCP server.

## What you share

Connecting grants the agent access to Jimdo on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Jira
Source: https://getskydive.com/docs/integrations/jira

Issue tracking and agile project management

## Example agent tasks

```text theme={null}
Connect Jira and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Jira.
```

```text theme={null}
Use the agent to interact with Jira directly from your conversation.
```

## Connecting

Connect Jira over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Jira.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Jira through the MCP server.

## What you share

Connecting grants the agent access to Jira on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Jotform
Source: https://getskydive.com/docs/integrations/jotform

Form builder with payments, apps and AI agents

## Example agent tasks

```text theme={null}
Connect Jotform and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Jotform.
```

```text theme={null}
Use the agent to interact with Jotform directly from your conversation.
```

## Connecting

Connect Jotform over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Jotform.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Jotform through the MCP server.

## What you share

Connecting grants the agent access to Jotform on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Kensho LLM-Ready API
Source: https://getskydive.com/docs/integrations/kensho

AI and data analytics for financial services

## Example agent tasks

```text theme={null}
Connect Kensho LLM-Ready API and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Kensho LLM-Ready API.
```

```text theme={null}
Use the agent to interact with Kensho LLM-Ready API directly from your conversation.
```

## Connecting

Connect Adaptive Retrieval over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Kensho LLM-Ready API.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Kensho LLM-Ready API through the MCP server.

## What you share

Connecting grants the agent access to Kensho LLM-Ready API on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Keploy Public
Source: https://getskydive.com/docs/integrations/keploy

Automated API testing from recorded traffic

## Example agent tasks

```text theme={null}
Connect Keploy Public and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Keploy Public.
```

```text theme={null}
Use the agent to interact with Keploy Public directly from your conversation.
```

## Connecting

Connect Keploy Public with an API key.

**1. Add the connection**

Generate an API key from your Keploy Public account settings.

**2. Authorize**

Copy the key. It is only shown once.

**3. The agent gets to work**

Paste it when prompted in the agent's connection card for Keploy Public.

## What you share

Connecting grants the agent access to Keploy Public on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Kestra MCP server
Source: https://getskydive.com/docs/integrations/kestra

Open source workflow orchestration platform

## What the agent can do

> **Search**
>
> Search Kestra plugins, docs, blueprints, or blogs.

> **Task Schema**
>
> Get the full schema/documentation for a specific Kestra task class.

> **List Plugins**
>
> List available Kestra plugins with summary info (name, title, group, description, categories, and counts per element typ

> **Plugin Tasks**
>
> List all tasks, triggers, and conditions for a specific Kestra plugin, grouped by subpackage.

## Example agent tasks

```text theme={null}
Search Kestra MCP server for relevant results and summarize what you find.
```

```text theme={null}
List everything available in Kestra MCP server and tell me what stands out.
```

```text theme={null}
Ask the agent to help you with Kestra MCP server tasks using natural language.
```

## Connecting

Connect Kestra over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Kestra.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Kestra through the MCP server.

## What you share

Connecting grants the agent access to Kestra MCP server on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Kit
Source: https://getskydive.com/docs/integrations/kit

Email marketing for creators, formerly ConvertKit

## Example agent tasks

```text theme={null}
Connect Kit and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Kit.
```

```text theme={null}
Use the agent to interact with Kit directly from your conversation.
```

## Connecting

Connect Kit Developer Docs over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Kit.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Kit through the MCP server.

## What you share

Connecting grants the agent access to Kit on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Klaviyo
Source: https://getskydive.com/docs/integrations/klaviyo

Email and SMS marketing for ecommerce

## Example agent tasks

```text theme={null}
Connect Klaviyo and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Klaviyo.
```

```text theme={null}
Use the agent to interact with Klaviyo directly from your conversation.
```

## Connecting

Connect Klaviyo over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Klaviyo.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Klaviyo through the MCP server.

## What you share

Connecting grants the agent access to Klaviyo on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Knock
Source: https://getskydive.com/docs/integrations/knock

Managed notification infrastructure across email, SMS and push

## Example agent tasks

```text theme={null}
Connect Knock and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Knock.
```

```text theme={null}
Use the agent to interact with Knock directly from your conversation.
```

## Connecting

Connect Knock with an API key.

**1. Add the connection**

Generate an API key from your Knock account settings.

**2. Authorize**

Copy the key. It is only shown once.

**3. The agent gets to work**

Paste it when prompted in the agent's connection card for Knock.

## What you share

Connecting grants the agent access to Knock on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Kraken MCP server
Source: https://getskydive.com/docs/integrations/kraken

Cryptocurrency exchange

## Example agent tasks

```text theme={null}
Connect Kraken MCP server and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Kraken MCP server.
```

```text theme={null}
Use the agent to interact with Kraken MCP server directly from your conversation.
```

## Connecting

Connect Kraken over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Kraken.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Kraken through the MCP server.

## What you share

Connecting grants the agent access to Kraken MCP server on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Krisp
Source: https://getskydive.com/docs/integrations/krisp

AI noise cancellation for calls and meetings

## Example agent tasks

```text theme={null}
Connect Krisp and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Krisp.
```

```text theme={null}
Use the agent to interact with Krisp directly from your conversation.
```

## Connecting

Connect Krisp over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Krisp.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Krisp through the MCP server.

## What you share

Connecting grants the agent access to Krisp on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Langfuse
Source: https://getskydive.com/docs/integrations/langfuse

LLM observability and tracing platform

## Example agent tasks

```text theme={null}
Connect Langfuse and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Langfuse.
```

```text theme={null}
Use the agent to interact with Langfuse directly from your conversation.
```

## Connecting

Connect Langfuse Docs over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Langfuse.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Langfuse through the MCP server.

## What you share

Connecting grants the agent access to Langfuse on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# HighLevel MCP (ChatGPT connector)
Source: https://getskydive.com/docs/integrations/leadconnectorhq

Marketing and CRM platform for agencies

## Example agent tasks

```text theme={null}
Connect HighLevel MCP (ChatGPT connector) and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from HighLevel MCP (ChatGPT connector).
```

```text theme={null}
Use the agent to interact with HighLevel MCP (ChatGPT connector) directly from your conversation.
```

## Connecting

Connect HighLevel (ChatGPT connector) over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add HighLevel MCP (ChatGPT connector).

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach HighLevel MCP (ChatGPT connector) through the MCP server.

## What you share

Connecting grants the agent access to HighLevel MCP (ChatGPT connector) on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# LeapScholar MCP Server
Source: https://getskydive.com/docs/integrations/leapscholar

Study abroad counselling and applications

## What the agent can do

> **Search Courses**
>
> Search university programs abroad.

## Example agent tasks

```text theme={null}
Search LeapScholar MCP Server for relevant results and summarize what you find.
```

```text theme={null}
Ask the agent to help you with LeapScholar MCP Server tasks using natural language.
```

## Connecting

Connect LeapScholar over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add LeapScholar.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach LeapScholar through the MCP server.

## What you share

Connecting grants the agent access to LeapScholar MCP Server on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Lemlist
Source: https://getskydive.com/docs/integrations/lemlist

Cold email outreach and sequencing

## Example agent tasks

```text theme={null}
Connect Lemlist and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Lemlist.
```

```text theme={null}
Use the agent to interact with Lemlist directly from your conversation.
```

## Connecting

Connect Lemlist over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Lemlist.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Lemlist through the MCP server.

## What you share

Connecting grants the agent access to Lemlist on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Lemon Squeezy
Source: https://getskydive.com/docs/integrations/lemonsqueezy

Merchant of record and payments for software

## Example agent tasks

```text theme={null}
Connect Lemon Squeezy and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Lemon Squeezy.
```

```text theme={null}
Use the agent to interact with Lemon Squeezy directly from your conversation.
```

## Connecting

Connect Lemon Squeezy with an API key.

**1. Add the connection**

Generate an API key from your Lemon Squeezy account settings.

**2. Authorize**

Copy the key. It is only shown once.

**3. The agent gets to work**

Paste it when prompted in the agent's connection card for Lemon Squeezy.

## What you share

Connecting grants the agent access to Lemon Squeezy on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Linear
Source: https://getskydive.com/docs/integrations/linear

Let your agent manage your Linear issues

Linear tracks software issues and projects. Connect it and your agent can read your issues and create or update them.

## Two ways to connect

Linear now runs a hosted MCP server. Connect it and the agent can work with your entire Linear workspace, not just the OAuth-scoped view.

* **Linear MCP (recommended).** Connect Linear's MCP server once and the agent can read and update issues, create projects, post comments, track cycle progress. See [Connect with MCP](#connect-with-mcp).
* **Linear OAuth.** The original sign-in integration, scoped to what you grant. See [Connect with OAuth](#connect-with-oauth).

## What the agent can do

> **Read issues**
>
> See issues by team, status, or assignee.

> **Create issues**
>
> File new issues.

> **Update issues**
>
> Change status, assignee, and priority.

> **See projects**
>
> Find the right project or team.

## Example agent tasks

```text theme={null}
When a customer hits a bug, file a Linear issue with the repro steps and tag the team.
```

```text theme={null}
Each morning, tell me what is in review and what is blocked on me.
```

```text theme={null}
Summarize what shipped this cycle for the changelog.
```

## Connect with MCP

**1. Add the Linear connection**

From the agent's connections, choose Linear. The platform connects you through Linear's MCP server.

**2. Sign in to Linear**

A connect card appears. Sign in to your Linear account and approve.

**3. The agent gets to work**

It can now read and update issues, create projects, post comments, track cycle progress through the MCP server.

> [!NOTE]
>
> MCP is the first-class connection when both options are available. See [Tools & MCP](/docs/capabilities/tools-and-mcp).

## Connect with OAuth

## What you share

During the connect flow you authorize what the agent can reach. It works only within what you grant. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

**1. Authorize Linear for the agent**

From the agent's connections, choose Linear and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[Linear API docs](https://developers.linear.app/docs)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# linkhut
Source: https://getskydive.com/docs/integrations/linkhut

Let your agent work with your linkhut bookmarks

linkhut is a social bookmarking service. Connect it and your agent can read and add to your saved links.

## What the agent can do

> **Read bookmarks**
>
> See your saved links.

> **Add bookmarks**
>
> Save new links with tags.

> **Browse tags**
>
> Find links by tag.

> **Search**
>
> Look up a saved link.

## Example agent tasks

```text theme={null}
Save this article to my linkhut with the right tags.
```

```text theme={null}
What have I bookmarked about pricing recently?
```

```text theme={null}
Pull my latest saved links into a reading list.
```

## What you share

During the connect flow you authorize what the agent can reach. It works only within what you grant. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Authorize linkhut for the agent**

From the agent's connections, choose linkhut and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[linkhut API docs](https://ln.ht/_/api)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Linkly
Source: https://getskydive.com/docs/integrations/linklyhq

Link tracking and URL shortening

## What the agent can do

> **Test Authentication**
>
> Test API Authentication

> **List Workspaces**
>
> Return details of authenticated workspace

> **Batch delete links**
>
> Batch delete multiple links

> **Update Workspace**
>
> Update workspace settings.

## Example agent tasks

```text theme={null}
List everything available in Linkly and tell me what stands out.
```

```text theme={null}
Update my Linkly records with the latest information.
```

```text theme={null}
Create a new entry in Linkly based on my notes.
```

## Connecting

Connect Linkly over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Linkly.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Linkly through the MCP server.

## What you share

Connecting grants the agent access to Linkly on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Linktree MCP Server
Source: https://getskydive.com/docs/integrations/linktr

Link-in-bio page for creators

## Example agent tasks

```text theme={null}
Connect Linktree MCP Server and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Linktree MCP Server.
```

```text theme={null}
Use the agent to interact with Linktree MCP Server directly from your conversation.
```

## Connecting

Connect Linktree over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Linktree.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Linktree through the MCP server.

## What you share

Connecting grants the agent access to Linktree MCP Server on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Laminar
Source: https://getskydive.com/docs/integrations/lmnr

Tracing and observability for LLM apps

## Example agent tasks

```text theme={null}
Connect Laminar and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Laminar.
```

```text theme={null}
Use the agent to interact with Laminar directly from your conversation.
```

## Connecting

Connect Laminar over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Laminar.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Laminar through the MCP server.

## What you share

Connecting grants the agent access to Laminar on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Local Falcon
Source: https://getskydive.com/docs/integrations/localfalcon

Local SEO rank tracking

## Example agent tasks

```text theme={null}
Connect Local Falcon and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Local Falcon.
```

```text theme={null}
Use the agent to interact with Local Falcon directly from your conversation.
```

## Connecting

Connect Local Falcon over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Local Falcon.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Local Falcon through the MCP server.

## What you share

Connecting grants the agent access to Local Falcon on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# LocationIQ
Source: https://getskydive.com/docs/integrations/locationiq

Geocoding and maps API

## Example agent tasks

```text theme={null}
Connect LocationIQ and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from LocationIQ.
```

```text theme={null}
Use the agent to interact with LocationIQ directly from your conversation.
```

## Connecting

Connect LocationIQ with an API key.

**1. Add the connection**

Generate an API key from your LocationIQ account settings.

**2. Authorize**

Copy the key. It is only shown once.

**3. The agent gets to work**

Paste it when prompted in the agent's connection card for LocationIQ.

## What you share

Connecting grants the agent access to LocationIQ on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# LogRocket
Source: https://getskydive.com/docs/integrations/logrocket

Session replay and frontend monitoring

## Example agent tasks

```text theme={null}
Connect LogRocket and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from LogRocket.
```

```text theme={null}
Use the agent to interact with LogRocket directly from your conversation.
```

## Connecting

Connect LogRocket over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add LogRocket.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach LogRocket through the MCP server.

## What you share

Connecting grants the agent access to LogRocket on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Loops
Source: https://getskydive.com/docs/integrations/loops-so

Let your agent manage contacts, send events, and deliver transactional email through Loops

Loops is a marketing and transactional email platform. Connect it and your agent can manage contacts, handle mailing lists, send events to trigger workflows, and deliver transactional email on your behalf.

## What the agent can do

> **Manage contacts**
>
> Find, create, and update contacts and their properties in your Loops account.

> **Work with mailing lists**
>
> Check membership status and add or remove contacts from your lists.

> **Send events**
>
> Fire events that update contact activity or trigger published Loops workflows.

> **Send transactional email**
>
> Deliver a published email template with the right data variables filled in.

## Example agent tasks

```text theme={null}
When a user completes onboarding, send them the welcome transactional email template.
```

```text theme={null}
Find all contacts tagged as churned and remove them from the active customers list.
```

```text theme={null}
Fire the "trial-ended" event for every user whose trial expires today.
```

## What you share

During the connect flow you authorize access to your Loops workspace. The agent works only within what you grant. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Add the Loops connection**

From the agent's connections, choose Loops. The platform connects you through the Loops MCP server.

**2. Sign in to Loops**

A connect card appears. Sign in to your Loops account and approve access.

**3. Choose your team**

If you belong to multiple Loops teams, the agent can list them and operate against the one you specify.

**4. The agent gets to work**

It can now manage contacts, fire events, handle mailing lists, and send transactional email.

## Learn more

> **[Loops docs](https://loops.so/docs)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Lucid
Source: https://getskydive.com/docs/integrations/lucid

Visual collaboration whiteboards and diagrams

## Example agent tasks

```text theme={null}
Connect Lucid and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Lucid.
```

```text theme={null}
Use the agent to interact with Lucid directly from your conversation.
```

## Connecting

Connect Lucid over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Lucid.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Lucid through the MCP server.

## What you share

Connecting grants the agent access to Lucid on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Lumin
Source: https://getskydive.com/docs/integrations/luminpdf

PDF editing and document signing

## Example agent tasks

```text theme={null}
Connect Lumin and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Lumin.
```

```text theme={null}
Use the agent to interact with Lumin directly from your conversation.
```

## Connecting

Connect Lumin over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Lumin.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Lumin through the MCP server.

## What you share

Connecting grants the agent access to Lumin on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Magneto Empleos MCP Server
Source: https://getskydive.com/docs/integrations/magneto365

Colombian online job board and recruiting

## What the agent can do

> **Ai Search Jobs**
>
> Intelligent job search using natural language.

> **Get Filters**
>
> Get available filters for job search on Magneto365: locations, categories, work modalities, experience, salary ranges, c

> **Get Job Detail**
>
> Get detailed information about a specific job listing on Magneto365 by its URL slug.

## Example agent tasks

```text theme={null}
Search Magneto Empleos MCP Server for relevant results and summarize what you find.
```

```text theme={null}
Pull the latest data from Magneto Empleos MCP Server and give me a summary.
```

## Connecting

Connect Magneto Empleos over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Magneto Empleos.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Magneto Empleos through the MCP server.

## What you share

Connecting grants the agent access to Magneto Empleos MCP Server on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Mailchimp
Source: https://getskydive.com/docs/integrations/mailchimp

Let your agent work with your Mailchimp audience and campaigns

Mailchimp runs email marketing and audiences. Connect it and your agent can read your lists and campaign results.

## What the agent can do

> **Read audiences**
>
> See your subscriber lists.

> **Read campaigns**
>
> Pull campaign details and stats.

> **See results**
>
> Read open and click rates.

> **Add subscribers**
>
> Add contacts to an audience.

## Example agent tasks

```text theme={null}
After each campaign sends, tell me how it did and what to try next time.
```

```text theme={null}
Every Monday, tell me which subscribers are going cold and suggest a win-back.
```

```text theme={null}
When someone signs up on our site, add them to the right audience automatically.
```

## What you share

During the connect flow you authorize what the agent can reach. It works only within what you grant. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Authorize Mailchimp for the agent**

From the agent's connections, choose Mailchimp and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[Mailchimp Marketing API docs](https://mailchimp.com/developer/marketing/api/)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Mailerlite
Source: https://getskydive.com/docs/integrations/mailerlite

Email marketing platform

## Example agent tasks

```text theme={null}
Connect Mailerlite and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Mailerlite.
```

```text theme={null}
Use the agent to interact with Mailerlite directly from your conversation.
```

## Connecting

Connect Mailerlite over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Mailerlite.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Mailerlite through the MCP server.

## What you share

Connecting grants the agent access to Mailerlite on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# MailerSend
Source: https://getskydive.com/docs/integrations/mailersend

Transactional email API

## Example agent tasks

```text theme={null}
Connect MailerSend and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from MailerSend.
```

```text theme={null}
Use the agent to interact with MailerSend directly from your conversation.
```

## Connecting

Connect MailerSend over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add MailerSend.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach MailerSend through the MCP server.

## What you share

Connecting grants the agent access to MailerSend on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Make
Source: https://getskydive.com/docs/integrations/make

Visual workflow automation platform

## Example agent tasks

```text theme={null}
Connect Make and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Make.
```

```text theme={null}
Use the agent to interact with Make directly from your conversation.
```

## Connecting

Connect Make over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Make.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Make through the MCP server.

## What you share

Connecting grants the agent access to Make on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Malwarebytes
Source: https://getskydive.com/docs/integrations/malwarebytes

Anti-malware and endpoint security

## Example agent tasks

```text theme={null}
Connect Malwarebytes and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Malwarebytes.
```

```text theme={null}
Use the agent to interact with Malwarebytes directly from your conversation.
```

## Connecting

Connect Malwarebytes over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Malwarebytes.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Malwarebytes through the MCP server.

## What you share

Connecting grants the agent access to Malwarebytes on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Manus
Source: https://getskydive.com/docs/integrations/manus

Autonomous AI agent for tasks and research

## Example agent tasks

```text theme={null}
Connect Manus and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Manus.
```

```text theme={null}
Use the agent to interact with Manus directly from your conversation.
```

## Connecting

Connect Manus over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Manus.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Manus through the MCP server.

## What you share

Connecting grants the agent access to Manus on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# ManyChat
Source: https://getskydive.com/docs/integrations/manychat

Chat marketing automation for Messenger, Instagram, WhatsApp, SMS and Telegram

ManyChat is chat marketing automation for Messenger, Instagram, WhatsApp, SMS, and Telegram. Connect it with a Personal Access Token and your agents can read subscriber info, tags, and custom fields, send messages, and manage tags and custom fields through the ManyChat Account Public API.

## What your agents can do

> **Look up subscribers**
>
> Pull subscriber info, tags, and custom fields by channel so agents can see who they are working with before acting.

> **Send messages and flows**
>
> Send a message or trigger a flow on Facebook, WhatsApp, or Instagram channels. Writes are confirmed with you first.

> **Manage tags and custom fields**
>
> Create, update, or remove tags and custom fields on a subscriber to keep audience segments current.

> **Inspect page setup**
>
> Read page info, existing tags, and custom fields so agents can map your ManyChat setup before making changes.

## Example agent tasks

```text theme={null}
Look up this subscriber's tags and custom fields in ManyChat and summarize their profile.
```

```text theme={null}
Add the subscribers from this list to our "VIP 2026" tag in ManyChat and confirm each write before running it.
```

```text theme={null}
Send the "Order shipped" flow to subscribers who tagged "order" in the last 24 hours, and report how many were sent.
```

## What you share

Connecting grants your agents access to the ManyChat Account Public API (api.manychat.com) using a Personal Access Token. Read-only endpoints like page info, tags, custom fields, and subscriber info are the default; sending messages or flows and creating or removing tags and custom fields are write actions, and agents confirm with you before executing them. Endpoints live under `/fb/` (and `/wa/` for WhatsApp, `/ig/` for Instagram). API access requires a paid ManyChat plan. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Generate an API token**

In ManyChat, go to Settings > API and click Generate your API Key. The token is shown once; copy it. API access requires a paid ManyChat plan.

**2. Add the ManyChat connection**

From your agent's connections, choose ManyChat. Paste your token into the connect card when prompted. The platform injects it as the `Authorization: Bearer` header automatically.

**3. Your agents get to work**

They can now read subscriber and page data and take confirmed write actions through the ManyChat API.

## Learn more

> **[Manychat API access guide](https://help.manychat.com/hc/en-us/articles/14959510331420-How-to-generate-a-token-for-the-Manychat-API-and-where-to-get-parameters)**
>
> How to generate an API token and where to find the API command reference.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Massive
Source: https://getskydive.com/docs/integrations/massive

Real-time and historical financial market data

## Example agent tasks

```text theme={null}
Connect Massive and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Massive.
```

```text theme={null}
Use the agent to interact with Massive directly from your conversation.
```

## Connecting

**MCP server**

Connect Massive over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Massive.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Massive through the MCP server.

**API key**

Connect Massive with an API key.

**1. Add the connection**

Generate an API key from your Massive account settings.

**2. Authorize**

Copy the key. It is only shown once.

**3. The agent gets to work**

Paste it when prompted in the agent's connection card for Massive.

## What you share

Connecting grants the agent access to Massive on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# MCP
Source: https://getskydive.com/docs/integrations/mcp

Purpose-built agent tools, from any MCP server

MCP (Model Context Protocol) is how services expose purpose-built tools directly to agents. When a service publishes an MCP server, connecting it gives your agent a set of tools designed for agent use, not just access to a REST API.

**One-click, via the catalog:**

Many MCP servers use OAuth with dynamic client registration, which means no manual setup. Click the service in the [catalog](/docs/integrations/catalog), authorize with your account, and the agent is connected. The credential is stored securely and attached to every request automatically.

**By URL, for any MCP server:**

A server your team hosts, or any server not in the catalog, connects the same way: point the agent at its URL.

```text theme={null}
Connect the MCP server at https://mcp.acme.com/mcp
```

Or add the URL in the agent's tools settings, or wire it up yourself in the agent's [repository config](/docs/agents/repo). The agent picks the server up on its next turn and the tools appear alongside the built-ins.

### When the server needs authorization

A Connect card opens. The agent follows the standard MCP authorization flow: it reads the server's OAuth discovery documents, registers as a client where the server allows it, and asks you to approve access in your browser. You approve once. After that the proxy attaches the token to every request automatically, and the token never appears in the agent's prompts. See [Connecting services](/docs/integrations/oauth) for how credentials are stored.

If you run the server, allow these two URLs in your OAuth setup:

| URL | Why |
| - | - |
| `https://www.skydive.com/api/v1/external-oauth/callback` | The redirect URI users land on after approving. Add it to your OAuth client's allowed redirects exactly as written; some identity providers compare it as an exact string. |
| `https://www.skydive.com/api/v1/external-oauth/client-metadata` | The client metadata document. Servers that register clients by URL fetch this document instead of taking a registration call, so it must stay publicly reachable and must not redirect. |

If your identity provider only allows pre-registered OAuth apps, tell the agent. A server that takes a static token connects without OAuth: the agent adds it to its [MCP config](/docs/agents/repo) with your key as a request header, and the proxy sends it on every request the same way.

> [!NOTE]
>
> An MCP connect card grants access at whatever level that server's OAuth app requests, which is often the full workspace or account. This is different from some OAuth flows (like Notion's) where you pick individual pages. On MCP Connect you won't see an item picker.

## What MCP adds over an API

An API gives the agent endpoints. An MCP server gives it tools with descriptions, typed arguments, and semantics written for an agent to read. The agent spends less effort working out how to call the service and more on the actual job.

## Managing a connection

Once a service is connected, manage it from the integration's card in your workspace. The card shows which agents use the connection and gives you two options to pull access back:

* **Remove from one agent.** The connection stays live for your other agents.
* **Disconnect entirely.** This revokes Skydive's access at the provider and removes it from every agent using it.

> **[Catalog](/docs/integrations/catalog)**
>
> Every service available to connect.

> **[Permissions](/docs/integrations/permissions)**
>
> How credentials stay out of the model's reach.


# MeetStream
Source: https://getskydive.com/docs/integrations/meetstream

Meeting bot API for Zoom, Google Meet, and Teams

MeetStream is a meeting bot API. Connect it and your agents can send a bot into Zoom, Google Meet, or Microsoft Teams meetings to record and transcribe them, then pull transcripts, participant media, chats, and speaker timelines. Bots can also be scheduled from a connected calendar or chatted into live meetings.

## What your agents can do

> **Send bots into meetings**
>
> Create bots for Zoom, Google Meet, and Teams meetings to record and transcribe them.

> **Pull transcripts and summaries**
>
> Retrieve finished transcripts, summaries, and per-participant media once a bot leaves.

> **Read and write meeting chats**
>
> Read chat and speaker timelines during or after a call, and send messages into live meetings.

> **Schedule bots from a calendar**
>
> Connect a calendar and have bots join recurring or scheduled meetings automatically.

## Example agent tasks

```text theme={null}
Send a bot to our 2pm standup and give me the transcript with action items when it ends.
```

```text theme={null}
Pull last week's sales call transcripts and summarize objections by participant.
```

```text theme={null}
Join the customer review meeting and post a summary of decisions into the chat before it ends.
```

## What you share

Connecting gives your agents access to your MeetStream account through its MCP server or REST API: they can create and manage bots, read transcripts, media, chats, and speaker timelines, and manage calendar scheduling. One thing to know: deleting a bot's data permanently removes its audio, video, and transcripts, and your agent should always confirm with you first. See [Permissioning](/docs/integrations/permissions).

## Connecting

**MCP server**

Connect MeetStream over MCP for direct access to its meeting bot tools. Authorization uses a MeetStream API key sent per request.

**1. Create an API key**

In the MeetStream dashboard open the API tab and create a key.

**2. Add the MeetStream connection**

From your agent's connections, choose MeetStream and paste your API key when prompted.

**3. Your agents get to work**

They can now send bots into meetings and pull transcripts through MeetStream's hosted MCP server at mcp.meetstream.ai.

**API key**

Connect MeetStream through its REST API for the full bot lifecycle.

**1. Create an API key**

In the MeetStream dashboard open the API tab and create a key.

**2. Add the MeetStream connection**

From your agent's connections, choose MeetStream and paste your API key when prompted.

**3. Your agents get to work**

Your agents can now call the MeetStream REST API at api.meetstream.ai, from creating bots to fetching transcripts.

## Learn more

> **[MeetStream API reference](https://docs.meetstream.ai)**
>
> Endpoint reference for bots, transcripts, media, and calendars.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Meetup MCP server
Source: https://getskydive.com/docs/integrations/meetup

Event and group organizing platform

## What the agent can do

> **Recommended Events In Person In Map**
>
> This is the default tool to use to search for in-person Meetup events.

## Example agent tasks

```text theme={null}
Ask the agent to help you with Meetup MCP server tasks using natural language.
```

## Connecting

Connect Meetup over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Meetup.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Meetup through the MCP server.

## What you share

Connecting grants the agent access to Meetup MCP server on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Mem
Source: https://getskydive.com/docs/integrations/mem

AI-powered note-taking app

## Example agent tasks

```text theme={null}
Connect Mem and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Mem.
```

```text theme={null}
Use the agent to interact with Mem directly from your conversation.
```

## Connecting

Connect Mem over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Mem.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Mem through the MCP server.

## What you share

Connecting grants the agent access to Mem on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Mercury
Source: https://getskydive.com/docs/integrations/mercury

Business banking for startups

## Example agent tasks

```text theme={null}
Connect Mercury and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Mercury.
```

```text theme={null}
Use the agent to interact with Mercury directly from your conversation.
```

## Connecting

Connect Mercury over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Mercury.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Mercury through the MCP server.

## What you share

Connecting grants the agent access to Mercury on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Mermaid MCP Server
Source: https://getskydive.com/docs/integrations/mermaid

Diagram generation from text

## Example agent tasks

```text theme={null}
Connect Mermaid MCP Server and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Mermaid MCP Server.
```

```text theme={null}
Use the agent to interact with Mermaid MCP Server directly from your conversation.
```

## Connecting

Connect Mermaid Chart Claude over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Mermaid.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Mermaid through the MCP server.

## What you share

Connecting grants the agent access to Mermaid MCP Server on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Metaview
Source: https://getskydive.com/docs/integrations/metaview

AI recruiting and interview intelligence

## Example agent tasks

```text theme={null}
Connect Metaview and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Metaview.
```

```text theme={null}
Use the agent to interact with Metaview directly from your conversation.
```

## Connecting

Connect Metaview over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Metaview.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Metaview through the MCP server.

## What you share

Connecting grants the agent access to Metaview on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Microsoft
Source: https://getskydive.com/docs/integrations/microsoft

Cloud services, productivity and identity platform

## Example agent tasks

```text theme={null}
Connect Microsoft and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Microsoft.
```

```text theme={null}
Use the agent to interact with Microsoft directly from your conversation.
```

## Connecting

**1. Authorize Microsoft for the agent**

From the agent's connections, choose Microsoft and sign in with your Microsoft account.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## What you share

Connecting grants the agent access to Microsoft on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[Microsoft Graph overview](https://learn.microsoft.com/en-us/graph/overview)**
>
> The API surface behind the Microsoft connection.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Midpage Legal Research
Source: https://getskydive.com/docs/integrations/midpage

AI legal research and drafting

## Example agent tasks

```text theme={null}
Connect Midpage Legal Research and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Midpage Legal Research.
```

```text theme={null}
Use the agent to interact with Midpage Legal Research directly from your conversation.
```

## Connecting

Connect Midpage Legal Research over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Midpage Legal Research.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Midpage Legal Research through the MCP server.

## What you share

Connecting grants the agent access to Midpage Legal Research on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Migma
Source: https://getskydive.com/docs/integrations/migma

AI email generation and marketing automation

## Example agent tasks

```text theme={null}
Connect Migma and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Migma.
```

```text theme={null}
Use the agent to interact with Migma directly from your conversation.
```

## Connecting

Connect Migma over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Migma.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Migma through the MCP server.

## What you share

Connecting grants the agent access to Migma on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Mintlify
Source: https://getskydive.com/docs/integrations/mintlify

Documentation platform for developer products

## Example agent tasks

```text theme={null}
Connect Mintlify and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Mintlify.
```

```text theme={null}
Use the agent to interact with Mintlify directly from your conversation.
```

## Connecting

Connect Mintlify with an API key.

**1. Add the connection**

Generate an API key from your Mintlify account settings.

**2. Authorize**

Copy the key. It is only shown once.

**3. The agent gets to work**

Paste it when prompted in the agent's connection card for Mintlify.

## What you share

Connecting grants the agent access to Mintlify on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Miro
Source: https://getskydive.com/docs/integrations/miro

Online whiteboard for visual collaboration

## Example agent tasks

```text theme={null}
Connect Miro and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Miro.
```

```text theme={null}
Use the agent to interact with Miro directly from your conversation.
```

## Connecting

Connect Miro over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Miro.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Miro through the MCP server.

## What you share

Connecting grants the agent access to Miro on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Mixedbread
Source: https://getskydive.com/docs/integrations/mixedbread

AI search and embeddings platform

## Example agent tasks

```text theme={null}
Connect Mixedbread and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Mixedbread.
```

```text theme={null}
Use the agent to interact with Mixedbread directly from your conversation.
```

## Connecting

Connect Mixedbread over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Mixedbread.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Mixedbread through the MCP server.

## What you share

Connecting grants the agent access to Mixedbread on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Mixmax
Source: https://getskydive.com/docs/integrations/mixmax

Sales engagement and email tracking

## Example agent tasks

```text theme={null}
Connect Mixmax and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Mixmax.
```

```text theme={null}
Use the agent to interact with Mixmax directly from your conversation.
```

## Connecting

Connect Mixmax over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Mixmax.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Mixmax through the MCP server.

## What you share

Connecting grants the agent access to Mixmax on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Mobbin
Source: https://getskydive.com/docs/integrations/mobbin

Design reference library with 600,000+ real product screens from top apps, searchable by your agents.

Mobbin is a design-reference library with over 600,000 real screens from shipped mobile and web products: onboarding flows, paywalls, checkout, KYC, empty states, settings, and more. Connect it and your agents can search those patterns directly, pulling relevant examples from real apps to inform design decisions without you having to browse manually.

Requires a Mobbin Pro, Team, or Enterprise plan.

## What your agents can do

> **Search UI patterns**
>
> Find real product screens by pattern type: paywalls, onboarding, checkout, empty states, and more.

> **Browse by app**
>
> Pull screens from specific apps to see how they handle a particular UI pattern.

> **Filter by platform**
>
> Narrow results to mobile or web products and find patterns suited to your platform.

> **Find dated references**
>
> Get screens with app name and publish date so your agents can cite current design patterns.

## Example agent tasks

```text theme={null}
Show me how top fintech apps handle their onboarding flow.
```

```text theme={null}
Find examples of paywall screens from productivity apps.
```

```text theme={null}
What does the checkout experience look like in the best e-commerce apps right now?
```

## What you share

Connecting gives your agents read access to Mobbin's screen library through your account. Your agents can search and retrieve screen images with app and pattern context. They cannot access private workspace content beyond what your Mobbin plan covers, and they cannot create, modify, or delete anything. Requires a Mobbin Pro, Team, or Enterprise plan. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Add the Mobbin connection**

From the agent's connections, choose Mobbin. The platform connects through Mobbin's MCP server.

**2. Sign in to Mobbin**

A connect card appears. Sign in to your Mobbin account and approve access. Your account must be on a paid plan.

**3. Your agents get to work**

They can now search Mobbin's full library of product screens to find design references on demand.

## Learn more

> **[Mobbin documentation](https://docs.mobbin.com/mcp/introduction)**
>
> How the Mobbin MCP server works and what your agents can do with it.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of your agents' reach.


# Monday
Source: https://getskydive.com/docs/integrations/monday

Work management and project tracking boards

## Example agent tasks

```text theme={null}
Connect Monday and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Monday.
```

```text theme={null}
Use the agent to interact with Monday directly from your conversation.
```

## Connecting

Connect Monday over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Monday.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Monday through the MCP server.

## What you share

Connecting grants the agent access to Monday on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# MongoDB Atlas App Services Admin
Source: https://getskydive.com/docs/integrations/mongodb

Backend and database services on MongoDB Atlas

## Example agent tasks

```text theme={null}
Connect MongoDB Atlas App Services Admin and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from MongoDB Atlas App Services Admin.
```

```text theme={null}
Use the agent to interact with MongoDB Atlas App Services Admin directly from your conversation.
```

## Connecting

Connect MongoDB Atlas App Services Admin with an API key.

**1. Add the connection**

Generate an API key from your MongoDB Atlas App Services Admin account settings.

**2. Authorize**

Copy the key. It is only shown once.

**3. The agent gets to work**

Paste it when prompted in the agent's connection card for MongoDB Atlas App Services Admin.

## What you share

Connecting grants the agent access to MongoDB Atlas App Services Admin on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# MyFonts MCP server
Source: https://getskydive.com/docs/integrations/monotype

Font marketplace and licensing

## Example agent tasks

```text theme={null}
Connect MyFonts MCP server and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from MyFonts MCP server.
```

```text theme={null}
Use the agent to interact with MyFonts MCP server directly from your conversation.
```

## Connecting

Connect MyFonts over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add MyFonts.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach MyFonts through the MCP server.

## What you share

Connecting grants the agent access to MyFonts MCP server on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Moov
Source: https://getskydive.com/docs/integrations/moov

Payment infrastructure for money movement

## Example agent tasks

```text theme={null}
Connect Moov and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Moov.
```

```text theme={null}
Use the agent to interact with Moov directly from your conversation.
```

## Connecting

Connect Moov with an API key.

**1. Add the connection**

Generate an API key from your Moov account settings.

**2. Authorize**

Copy the key. It is only shown once.

**3. The agent gets to work**

Paste it when prompted in the agent's connection card for Moov.

## What you share

Connecting grants the agent access to Moov on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Morningstar
Source: https://getskydive.com/docs/integrations/morningstar

Investment research and ratings

## Example agent tasks

```text theme={null}
Connect Morningstar and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Morningstar.
```

```text theme={null}
Use the agent to interact with Morningstar directly from your conversation.
```

## Connecting

Connect Morningstar over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Morningstar.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Morningstar through the MCP server.

## What you share

Connecting grants the agent access to Morningstar on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# MotherDuck
Source: https://getskydive.com/docs/integrations/motherduck

Serverless cloud analytics built on DuckDB

## Example agent tasks

```text theme={null}
Connect MotherDuck and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from MotherDuck.
```

```text theme={null}
Use the agent to interact with MotherDuck directly from your conversation.
```

## Connecting

Connect MotherDuck over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add MotherDuck.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach MotherDuck through the MCP server.

## What you share

Connecting grants the agent access to MotherDuck on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Motion Creative Analytics
Source: https://getskydive.com/docs/integrations/motionapp

Ad creative performance analytics

## Example agent tasks

```text theme={null}
Connect Motion Creative Analytics and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Motion Creative Analytics.
```

```text theme={null}
Use the agent to interact with Motion Creative Analytics directly from your conversation.
```

## Connecting

Connect Motion Creative Analytics over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Motion Creative Analytics.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Motion Creative Analytics through the MCP server.

## What you share

Connecting grants the agent access to Motion Creative Analytics on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Natural
Source: https://getskydive.com/docs/integrations/natural

Payments and card infrastructure for AI agents

## Example agent tasks

```text theme={null}
Connect Natural and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Natural.
```

```text theme={null}
Use the agent to interact with Natural directly from your conversation.
```

## Connecting

Connect Natural over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Natural.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Natural through the MCP server.

## What you share

Connecting grants the agent access to Natural on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Ubersuggest MCP Server
Source: https://getskydive.com/docs/integrations/neilpatelapi

SEO keyword research and analysis

## Example agent tasks

```text theme={null}
Connect Ubersuggest MCP Server and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Ubersuggest MCP Server.
```

```text theme={null}
Use the agent to interact with Ubersuggest MCP Server directly from your conversation.
```

## Connecting

Connect Ubersuggest over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Ubersuggest.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Ubersuggest through the MCP server.

## What you share

Connecting grants the agent access to Ubersuggest MCP Server on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Neon
Source: https://getskydive.com/docs/integrations/neon

Serverless Postgres database

## Example agent tasks

```text theme={null}
Connect Neon and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Neon.
```

```text theme={null}
Use the agent to interact with Neon directly from your conversation.
```

## Connecting

Connect Neon over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Neon.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Neon through the MCP server.

## What you share

Connecting grants the agent access to Neon on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Netlify
Source: https://getskydive.com/docs/integrations/netlify

Deployment platform with first-class support for any framework

## Example agent tasks

```text theme={null}
Connect Netlify and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Netlify.
```

```text theme={null}
Use the agent to interact with Netlify directly from your conversation.
```

## Connecting

**MCP server**

Connect Netlify over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Netlify.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Netlify through the MCP server.

**API key**

Connect Netlify with an API key.

**1. Add the connection**

Generate an API key from your Netlify account settings.

**2. Authorize**

Copy the key. It is only shown once.

**3. The agent gets to work**

Paste it when prompted in the agent's connection card for Netlify.

## What you share

Connecting grants the agent access to Netlify on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Neurnav
Source: https://getskydive.com/docs/integrations/neurnav

Version control for documents with AI review tracking

## Example agent tasks

```text theme={null}
Connect Neurnav and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Neurnav.
```

```text theme={null}
Use the agent to interact with Neurnav directly from your conversation.
```

## Connecting

Connect Neurnav over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Neurnav.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Neurnav through the MCP server.

## What you share

Connecting grants the agent access to Neurnav on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# New Relic
Source: https://getskydive.com/docs/integrations/newrelic

Application performance monitoring

## Example agent tasks

```text theme={null}
Connect New Relic and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from New Relic.
```

```text theme={null}
Use the agent to interact with New Relic directly from your conversation.
```

## Connecting

Connect New Relic over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add New Relic.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach New Relic through the MCP server.

## What you share

Connecting grants the agent access to New Relic on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Notion
Source: https://getskydive.com/docs/integrations/notion

Let your agent read and update your Notion workspace

Notion is where a lot of teams keep their notes, docs, and trackers. Connect it and your agent can read those pages and databases, pull what it needs, and write updates back.

## Two ways to connect

You can connect Notion two ways. For most people, **Notion's MCP server is the smoother option**, especially if you want the agent to reach your whole workspace.

* **Notion MCP (recommended).** Connect Notion's hosted MCP server once and the agent works across your Notion through it. This avoids the page-by-page granting that trips people up with the integration, and it's the path we've seen work most reliably for broad access. See [Connect with MCP](#connect-with-mcp).
* **The Notion integration (Add button).** A one-click OAuth connection scoped to the specific pages you pick. Good for a tight, single-area grant. See [Connect the integration](#connecting-the-integration).

> [!TIP]
>
> If you want the agent to keep up with your whole Notion, or you've hit "the agent can't see my pages" with the integration, use MCP.

## What the agent can do

> **Read pages and notes**
>
> Pull the content of any page you've shared, like meeting notes or a project brief.

> **Work with databases**
>
> Query a database, filter and sort rows, and read entry properties.

> **Create and update**
>
> Add new pages, append to existing ones, and update database entries.

> **Search the workspace**
>
> Find the right page or database by keyword before acting on it.

## Example agent tasks

```text theme={null}
Every Monday, pull the past week's meeting notes into a single recap page and share it with the team.
```

```text theme={null}
When I approve a draft in my Content Calendar, flip its status to Scheduled and set the publish date.
```

```text theme={null}
Read the Onboarding page, turn the checklist into tasks, and DM the new hire their first day plan.
```

## Connect with MCP

Notion runs a hosted [MCP](/docs/capabilities/tools-and-mcp) server, and connecting through it is the simplest way to give the agent broad, reliable access to your Notion.

**1. Add Notion's MCP server**

Point the agent at Notion's MCP server, `https://mcp.notion.com/mcp`. Ask the agent to connect it, or add it from the agent's tools.

**2. Authorize once**

A connect card appears. Sign in to Notion and approve. Skydive stores the credential securely and its [proxy](/docs/integrations/permissions) attaches it to every request from then on.

**3. The agent gets to work**

It can now read and act across your Notion through the MCP server, no page-by-page granting to maintain.

> [!NOTE]
>
> MCP is one connection for your workspace, rather than the integration's per-page grants. It's the better default when you want the agent to work across your Notion. See [Tools & MCP](/docs/capabilities/tools-and-mcp).

## What you share

The rest of this page covers the **Notion integration** (the one-click Add option). If you connected through MCP above, that flow handles access for you.

With the integration, Notion access is page by page. During the connect flow you pick exactly which pages and databases the agent can see, and it can't touch anything you didn't share. Grant a single project page for a focused task, or a top-level page to give the agent a whole area to work in.

The connection is **live, not a one-time snapshot**. The agent reads Notion fresh every time, so it always sees the current content, including edits made after you connected. What it can reach is set by *which pages you grant*, not *when* you connected.

Two things trip people up:

* **It only sees what you selected.** If the agent seems to be missing your existing pages, they most likely weren't picked in Notion's page picker when you connected. It's a scope gap, not a sync delay.
* **New pages aren't added automatically.** A page you create later isn't covered unless it lives under a page you already granted, or you share it with the connection.

### Give the agent access to more pages

To widen what the agent can reach, do either of these in Notion:

* **Re-run the connection** and select more pages in the page picker, or
* Open a page's \`\` menu and choose **Connections** to add the agent to that page.

Once a page is granted, the agent reads and writes it live from then on.

### Check and manage access from a page

Open any page's \`\` menu (top right) and go to **Connections** to see and control the agent's access to that page:

* **Active connections** lists the connections on the page. Your agent appears here (for example, **Skydive**), showing who authorized it.
* **Capabilities** shows exactly what the connection can do on the page: read content, insert content, update content, comment, and read comments.
* **Add connection** grants the agent access to this page if it isn't already connected.
* **Manage connections** opens the workspace connection settings.
* **Disconnect** removes the agent's access to that page.

![The Connections panel on a Notion page, listing the Skydive connection under Active connections with its capabilities: can read, insert, and update content, comment, and read comments](https://mintcdn.com/skydive/dydnCUduqRsFQYy1/images/integrations/notion-connections-menu.png?fit=max\&auto=format\&n=dydnCUduqRsFQYy1\&q=85\&s=47e41f76a572af7c9f3efa03b12c6752)

A Notion page's Connections menu, showing the agent's active connection and its capabilities

### Keep new content and pages flowing in

Edits are always picked up: because access is live, changes to a page the agent already has need nothing from you.

New pages are the part to plan for. The cleanest way to keep them included is to **grant a parent page or a whole database** instead of individual pages. In Notion's own words, giving a connection access to a parent page provides access to its child pages, so pages you add underneath it are reachable without re-granting each one. Sharing a single leaf page only shares that page, so new sibling pages won't be seen.

* **Want the agent to keep up with an area?** Grant the top-level page or database for that area, and keep new work inside it.
* **A new page isn't being picked up?** It's outside a granted page. Open its \`\` menu, choose **Connections**, and add the agent, or move it under a page the agent already has.

## Connecting the integration

**1. Authorize Notion for the agent**

From the agent's connections, choose Notion and sign in.

**2. Pick the pages to share**

Use Notion's page picker to grant access to the pages and databases the agent should work with. Select the existing pages you want it to use here; anything you don't pick stays invisible to the agent.

**3. The agent gets to work**

It can now read and update everything you shared, through the secure [proxy](/docs/integrations/permissions).

> [!TIP]
>
> Granting a top-level page also covers the pages nested under it, so pick a parent page to give the agent a whole area in one grant instead of adding pages one at a time.

## Learn more

> **[Notion API docs](https://developers.notion.com/docs/getting-started)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Novu
Source: https://getskydive.com/docs/integrations/novu

Open-source notification infrastructure you can self-host

## Example agent tasks

```text theme={null}
Connect Novu and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Novu.
```

```text theme={null}
Use the agent to interact with Novu directly from your conversation.
```

## Connecting

Connect Novu over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Novu.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Novu through the MCP server.

## What you share

Connecting grants the agent access to Novu on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# NOWPayments
Source: https://getskydive.com/docs/integrations/nowpayments

Cryptocurrency payment gateway

## Example agent tasks

```text theme={null}
Connect NOWPayments and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from NOWPayments.
```

```text theme={null}
Use the agent to interact with NOWPayments directly from your conversation.
```

## Connecting

Connect NOWPayments with an API key.

**1. Add the connection**

Generate an API key from your NOWPayments account settings.

**2. Authorize**

Copy the key. It is only shown once.

**3. The agent gets to work**

Paste it when prompted in the agent's connection card for NOWPayments.

## What you share

Connecting grants the agent access to NOWPayments on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Nuxt UI MCP Server
Source: https://getskydive.com/docs/integrations/nuxt

Vue.js UI component library

## Example agent tasks

```text theme={null}
Connect Nuxt UI MCP Server and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Nuxt UI MCP Server.
```

```text theme={null}
Use the agent to interact with Nuxt UI MCP Server directly from your conversation.
```

## Connecting

Connect Nuxt UI over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Nuxt UI.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Nuxt UI through the MCP server.

## What you share

Connecting grants the agent access to Nuxt UI MCP Server on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# OAuth
Source: https://getskydive.com/docs/integrations/oauth

One-click authorization for services that support it

For services with one-click OAuth, connecting is a single authorize and you are done. Skydive handles token refresh automatically, so the connection keeps working without re-authorizing.

This is the preferred method for account-level access when the service supports it. Check the [catalog](/docs/integrations/catalog) to see which method a service offers.

For services with one-click OAuth, connecting is a single authorize and you're done. The platform handles token refresh automatically, so the connection keeps working without re-authorizing.

**1. The agent requests a connection**

Either you set it up from the catalog ahead of time, or the agent prompts you with a connect card when it hits a task that needs it.

**2. A connect card appears**

Click it, sign in with the provider, and approve the scopes.

**3. You're connected**

The agent can now act in that service securely, through the proxy.

## Per-item grants

Some providers scope the grant to items you pick on the consent screen. Notion is the clearest case: the token sees only the pages you selected, plus their children.

With one of these, an empty search result usually means "not shared with this connection" rather than "does not exist". Two fixes: share the item with the connection inside the provider, or re-run the connection to reopen the consent screen and widen the selection. Re-running renews the connection in place, and the existing grant keeps working until the new authorization completes.

## Managing a connection

Once a service is connected, manage it from the integration's card in your workspace. The card shows which agents use the connection and gives you two options to pull access back:

* **Remove from one agent.** The connection stays live for your other agents.
* **Disconnect entirely.** This revokes Skydive's access at the provider and removes it from every agent using it.

> **[Catalog](/docs/integrations/catalog)**
>
> Every service available to connect.

> **[Permissions](/docs/integrations/permissions)**
>
> How credentials stay out of the model's reach.


# OKX MCP chat endpoint
Source: https://getskydive.com/docs/integrations/okx

Cryptocurrency exchange and trading

## What the agent can do

> **Gateway.token-comparison**
>
> Compare multiple cryptocurrencies side by side.

> **Gateway.crypto-rank**
>
> Query cryptocurrency up/down rank list.

> **Gateway.crypto-chart**
>
> Display cryptocurrency price charts and market information.

## Example agent tasks

```text theme={null}
Ask the agent to help you with OKX MCP chat endpoint tasks using natural language.
```

## Connecting

Connect OKX chat endpoint over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add OKX MCP chat endpoint.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach OKX MCP chat endpoint through the MCP server.

## What you share

Connecting grants the agent access to OKX MCP chat endpoint on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Imovirtual MCP Server
Source: https://getskydive.com/docs/integrations/olx

Polish real estate listings

## Example agent tasks

```text theme={null}
Connect Imovirtual MCP Server and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Imovirtual MCP Server.
```

```text theme={null}
Use the agent to interact with Imovirtual MCP Server directly from your conversation.
```

## Connecting

Connect Imovirtual over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Imovirtual.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Imovirtual through the MCP server.

## What you share

Connecting grants the agent access to Imovirtual MCP Server on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Footprint
Source: https://getskydive.com/docs/integrations/onefootprint

Identity verification and KYC infrastructure

## Example agent tasks

```text theme={null}
Connect Footprint and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Footprint.
```

```text theme={null}
Use the agent to interact with Footprint directly from your conversation.
```

## Connecting

Connect Footprint with an API key.

**1. Add the connection**

Generate an API key from your Footprint account settings.

**2. Authorize**

Copy the key. It is only shown once.

**3. The agent gets to work**

Paste it when prompted in the agent's connection card for Footprint.

## What you share

Connecting grants the agent access to Footprint on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# OpenBudget
Source: https://getskydive.com/docs/integrations/openbudget

AI assistant for bank accounts and budgeting

## Example agent tasks

```text theme={null}
Connect OpenBudget and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from OpenBudget.
```

```text theme={null}
Use the agent to interact with OpenBudget directly from your conversation.
```

## Connecting

Connect OpenBudget over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add OpenBudget.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach OpenBudget through the MCP server.

## What you share

Connecting grants the agent access to OpenBudget on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# OpenRouter
Source: https://getskydive.com/docs/integrations/openrouter

Unified API access to multiple AI models

## Example agent tasks

```text theme={null}
Connect OpenRouter and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from OpenRouter.
```

```text theme={null}
Use the agent to interact with OpenRouter directly from your conversation.
```

## Connecting

Connect OpenRouter over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add OpenRouter.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach OpenRouter through the MCP server.

## What you share

Connecting grants the agent access to OpenRouter on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# OpenStatus
Source: https://getskydive.com/docs/integrations/openstatus

Open source uptime and status monitoring

## Example agent tasks

```text theme={null}
Connect OpenStatus and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from OpenStatus.
```

```text theme={null}
Use the agent to interact with OpenStatus directly from your conversation.
```

## Connecting

Connect OpenStatus over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add OpenStatus.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach OpenStatus through the MCP server.

## What you share

Connecting grants the agent access to OpenStatus on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Optimizely Experimentation Data Authentication
Source: https://getskydive.com/docs/integrations/optimizely

A/B testing and experimentation platform

## Example agent tasks

```text theme={null}
Connect Optimizely Experimentation Data Authentication and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Optimizely Experimentation Data Authentication.
```

```text theme={null}
Use the agent to interact with Optimizely Experimentation Data Authentication directly from your conversation.
```

## Connecting

Connect Optimizely Experimentation Data Authentication with an API key.

**1. Add the connection**

Generate an API key from your Optimizely Experimentation Data Authentication account settings.

**2. Authorize**

Copy the key. It is only shown once.

**3. The agent gets to work**

Paste it when prompted in the agent's connection card for Optimizely Experimentation Data Authentication.

## What you share

Connecting grants the agent access to Optimizely Experimentation Data Authentication on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Otter
Source: https://getskydive.com/docs/integrations/otter

Meeting transcription and AI notetaking

## Example agent tasks

```text theme={null}
Connect Otter and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Otter.
```

```text theme={null}
Use the agent to interact with Otter directly from your conversation.
```

## Connecting

**MCP server**

Connect Otter over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Otter.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Otter through the MCP server.

**Enterprise API key**

Connect Otter Public API with an Enterprise API key.

**1. Get an API key**

In Otter, open Integrations, then Developer, and create an API key. Requires an Otter Enterprise workspace.

**2. Add the connection**

From the agent's connections, choose Otter and paste your Enterprise API key when prompted.

**3. The agent gets to work**

The agent can now reach the Otter API on your behalf.

## What you share

Connecting grants the agent access to Otter on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# OutCallerAI
Source: https://getskydive.com/docs/integrations/outcallerai

AI voice agents for outbound sales calls

## Example agent tasks

```text theme={null}
Connect OutCallerAI and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from OutCallerAI.
```

```text theme={null}
Use the agent to interact with OutCallerAI directly from your conversation.
```

## Connecting

Connect OutCallerAI with an API key.

**1. Add the connection**

Generate an API key from your OutCallerAI account settings.

**2. Authorize**

Copy the key. It is only shown once.

**3. The agent gets to work**

Paste it when prompted in the agent's connection card for OutCallerAI.

## What you share

Connecting grants the agent access to OutCallerAI on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Outlook
Source: https://getskydive.com/docs/integrations/outlook

Let your agent read and send email, manage your calendar, and work with contacts in Outlook

Outlook is your email and calendar. Connect it and your agent can read your mail, draft and send messages, manage your schedule, and look up contacts on your behalf.

## What your agents can do

> **Read mail**
>
> Catch up on your inbox and pull a thread.

> **Draft and send**
>
> Prepare replies and send messages.

> **Manage your calendar**
>
> Read events, schedule meetings, and move things around.

> **Look up contacts**
>
> Find people and their details from your contacts.

## Example agent tasks

```text theme={null}
Every morning, triage my inbox: summarize what matters, draft replies to the routine ones, and flag what's urgent.
```

```text theme={null}
Find a 30 minute slot with the team this week and send the invite from my calendar.
```

```text theme={null}
When a contract lands in my inbox, pull the key dates into my calendar and draft a reply.
```

## What you share

You sign in with your Microsoft account and grant access to your Outlook mail, calendar, and contacts. The agent works only within what you grant. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Authorize Outlook for the agent**

From the agent's connections, choose Outlook and sign in with your Microsoft account.

**2. Approve the access**

Confirm what the agent can reach: mail, calendar, and contacts. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[Outlook](https://outlook.live.com/owa/)**
>
> The product itself.

> **[Microsoft Graph docs](https://learn.microsoft.com/en-us/microsoft-365/)**
>
> The API surface your agent works through.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Outreach
Source: https://getskydive.com/docs/integrations/outreach

Sales engagement and outreach platform

## Example agent tasks

```text theme={null}
Connect Outreach and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Outreach.
```

```text theme={null}
Use the agent to interact with Outreach directly from your conversation.
```

## Connecting

Connect Outreach over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Outreach.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Outreach through the MCP server.

## What you share

Connecting grants the agent access to Outreach on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Overview
Source: https://getskydive.com/docs/integrations/overview

Your agent connects the services it needs. You approve with one click

To do real work, an agent needs access to your tools: your GitHub, your Google account, an API, an internal system. You don't have to set any of it up in advance. **Your agent connects the services it needs, when it needs them. You approve with one click.**

## Your agent does the setup

When an agent reaches a task that needs access it doesn't have, it asks for exactly what it needs, right then. You approve, and it carries on.

**1. The agent needs access**

It hits a task requiring a service it isn't connected to.

**2. You approve**

A connect card appears in the conversation. One click signs you in or accepts the key.

**3. The agent continues**

With access in place, it finishes the job, using the credential securely through the proxy.

## The agent never sees your secrets

When you connect a service, the credential is stored securely and injected at the **network layer**: a proxy attaches it to the agent's outbound requests on the wire. The agent acts on your behalf, but the raw token never appears in its prompts, its logs, or its memory.

You grant real access without handing the model your keys.

## Ways to connect

Most of the time the agent prompts you and you approve. To set something up yourself:

> **[Connect a service](/docs/integrations/oauth)**
>
> One-click authorization for supported services, with tokens refreshed automatically.

> **[Supported integrations](/docs/integrations/catalog)**
>
> See the services your agent can connect to today.

> **[API keys & secrets](/docs/integrations/secrets)**
>
> Give the agent a key for a service, stored securely.

> **[GitHub](/docs/integrations/github)**
>
> Give your agent its own GitHub identity to read and write code.

> **[Permissioning](/docs/integrations/permissions)**
>
> How credentials are scoped and injected without ever leaking.

> **[Connect your first service](/docs/integrations/oauth)**
>
> Walk through the authorization flow.


# Helium 10
Source: https://getskydive.com/docs/integrations/pacvue

Amazon, Walmart and TikTok Shop seller tools

## Example agent tasks

```text theme={null}
Connect Helium 10 and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Helium 10.
```

```text theme={null}
Use the agent to interact with Helium 10 directly from your conversation.
```

## Connecting

Connect Helium 10 over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Helium 10.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Helium 10 through the MCP server.

## What you share

Connecting grants the agent access to Helium 10 on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Paddle
Source: https://getskydive.com/docs/integrations/paddle

Payments and billing infrastructure for software companies

## Example agent tasks

```text theme={null}
Connect Paddle and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Paddle.
```

```text theme={null}
Use the agent to interact with Paddle directly from your conversation.
```

## Connecting

Connect Paddle docs over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Paddle.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Paddle through the MCP server.

## What you share

Connecting grants the agent access to Paddle on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# PagerDuty
Source: https://getskydive.com/docs/integrations/pagerduty

Let your agent work with your PagerDuty incidents

PagerDuty handles on-call and incidents. Connect it and your agent can read incidents, schedules, and help manage responses.

## What the agent can do

> **Read incidents**
>
> See open and recent incidents.

> **See on-call**
>
> Check who's on call.

> **Manage incidents**
>
> Acknowledge, resolve, and note incidents.

> **Schedules**
>
> Read on-call schedules.

## Example agent tasks

```text theme={null}
When an incident fires, gather the context and post a clear summary to the channel.
```

```text theme={null}
Each morning, tell me what broke overnight and whether it's resolved.
```

```text theme={null}
Summarize this week's incidents and the patterns worth fixing.
```

## What you share

During the connect flow you authorize what the agent can reach. It works only within what you grant. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Authorize PagerDuty for the agent**

From the agent's connections, choose PagerDuty and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[PagerDuty API docs](https://developer.pagerduty.com/api-reference/)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Pandadoc
Source: https://getskydive.com/docs/integrations/pandadoc

Document creation, e-signature and proposals

## Example agent tasks

```text theme={null}
Connect Pandadoc and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Pandadoc.
```

```text theme={null}
Use the agent to interact with Pandadoc directly from your conversation.
```

## Connecting

Connect Pandadoc over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Pandadoc.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Pandadoc through the MCP server.

## What you share

Connecting grants the agent access to Pandadoc on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Paper
Source: https://getskydive.com/docs/integrations/paper

Let your agent read and edit the design open in Paper Desktop

Paper is a design tool with a local MCP server for the file open in Paper Desktop. Skydive reaches it through [Portal](/docs/integrations/portal), so your agent can work directly on the canvas.

## What your agent can do

> **Inspect the design**
>
> Read Paper layers and styles.

> **Build on the canvas**
>
> Create editable Paper artboards.

> **Update existing work**
>
> Change content, layout, and styling.

> **Export assets**
>
> Export designs in common formats.

## Example agent tasks

```text theme={null}
Build a settings screen in the open Paper file using the design tokens already on the canvas.
```

```text theme={null}
Update the selected pricing card with this copy and keep the existing layout.
```

```text theme={null}
Export the selected logo as SVG and the hero artboard as a 2x PNG.
```

## What you share

Paper's MCP server works with the file currently open in Paper Desktop. It can read the design and make live changes when you ask, including moving, restyling, or deleting layers.

Paper uses [Portal](/docs/integrations/portal), so connecting it grants that agent access to your computer, not only Paper. Portal is off until you approve it for the agent, and you can revoke access at any time. No Paper password or token is shared.

## Connecting

**1. Install the desktop apps**

Install Paper Desktop and [Skydive Desktop](/docs/desktop/mac). You can use the [Skydive CLI](/docs/cli/overview) instead of Skydive Desktop.

**2. Open a Paper file**

Open the file you want the agent to use. Paper starts its local MCP server while the file is open.

**3. Choose an agent**

Open **Integrations** in Skydive and choose **Paper**. Select an agent, then click **Go to chat**.

**4. Approve Portal access**

Approve the agent's request in Skydive Desktop. The agent finds Paper's local server and completes the connection.

## Learn more

> **[Paper MCP docs](https://paper.design/docs/mcp)**
>
> Paper's setup and tool reference.

> **[Portal](/docs/integrations/portal)**
>
> How computer access works.


# Parker
Source: https://getskydive.com/docs/integrations/parker

AI creative strategist for performance marketing - connect your ad data, reviews, and competitor research

Parker is an AI creative strategist for performance marketing teams and DTC brands. It connects to your Meta ad account, customer reviews, post-purchase surveys, and competitor brands, tagging and indexing everything so your agents can search across your live creative data without manual exports.

## What your agents can do

> **Search your ad library**
>
> Find ads by format, emotion, awareness level, or any tag across your Meta account and the competitor brands you track.

> **Pull customer reviews and surveys**
>
> Retrieve review and post-purchase survey data so agents can surface your customers' real language for briefs and copy.

> **Research competitor creative**
>
> Access tagged competitor ads, Reddit discussions, TikTok content, and organic performance to find what's working in your market.

> **Surface attribution data**
>
> Query ad performance and Northbeam attribution data to ground creative decisions in actual numbers.

## Example agent tasks

```text theme={null}
Pull every testimonial-style ad my competitors ran this month and summarize what's working.
```

```text theme={null}
Find my top-performing hooks from the last 90 days and write three variations for a new test.
```

```text theme={null}
What do customers say most often about our checkout experience? Pull it from reviews and surveys.
```

## What you share

Connecting gives your agents read-only access to your Parker brand workspace: Meta ad creative and performance data, customer reviews, post-purchase surveys, competitor ad libraries, and attribution data from connected sources. Most tools are read-only. The `brand_discovery` tool can update brand settings. Requires an active paid Parker subscription with a brand workspace and Meta ad account connected. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Add the Parker connection**

From the agent's connections, choose Parker. The platform connects through Parker's MCP server.

**2. Sign in to Parker**

A connect card appears. Sign in with your Parker account at heyparker.ai and approve access. You need an active paid subscription and a brand workspace with your Meta ad account connected.

**3. Your agents get to work**

They can now search your ad library, pull reviews and competitor data, and surface attribution insights from your Parker workspace.

## Learn more

> **[Parker docs](https://docs.heyparker.ai)**
>
> Parker's MCP documentation and getting-started guides.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Passage
Source: https://getskydive.com/docs/integrations/passage

Compliance access layer for onchain capital markets

## Example agent tasks

```text theme={null}
Connect Passage and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Passage.
```

```text theme={null}
Use the agent to interact with Passage directly from your conversation.
```

## Connecting

Connect Passage with an API key.

**1. Add the connection**

Generate an API key from your Passage account settings.

**2. Authorize**

Copy the key. It is only shown once.

**3. The agent gets to work**

Paste it when prompted in the agent's connection card for Passage.

## What you share

Connecting grants the agent access to Passage on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# PayPal
Source: https://getskydive.com/docs/integrations/paypal

Let your agent work with your PayPal payments

PayPal handles online payments. Connect it and your agent can create invoices, pull transaction data, and report on your account activity.

## Two ways to connect

PayPal now runs a hosted MCP server. Connect it and the agent can work with your PayPal account to handle invoicing and reporting.

* **PayPal MCP (recommended).** Connect PayPal's MCP server once and the agent can create and send invoices, pull transaction reports, check account activity. See [Connect with MCP](#connect-with-mcp).
* **PayPal OAuth.** The original sign-in integration, scoped to what you grant. See [Connect with OAuth](#connect-with-oauth).

## What the agent can do

> **Create invoices**
>
> Draft and send invoices to customers.

> **Pull transactions**
>
> Retrieve payment and transaction history.

> **Check account activity**
>
> Get a summary of recent sales and payouts.

> **Manage payment flows**
>
> Work with orders and payment processing.

## Example agent tasks

```text theme={null}
Create and send invoices to everyone on this list from my spreadsheet.
```

```text theme={null}
Give me a summary of transactions from the past 30 days.
```

```text theme={null}
Which of my customers have unpaid invoices?
```

## Connect with MCP

**1. Add the PayPal connection**

From the agent's connections, choose PayPal. The platform connects you through PayPal's MCP server.

**2. Sign in to PayPal**

A connect card appears. Sign in to your PayPal account and approve.

**3. The agent gets to work**

It can now create and send invoices, pull transaction reports, check account activity through the MCP server.

> [!NOTE]
>
> MCP is the first-class connection when both options are available. See [Tools & MCP](/docs/capabilities/tools-and-mcp).

## Connect with OAuth

## What you share

During the connect flow you authorize what the agent can reach. It works only within what you grant. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

**1. Authorize PayPal for the agent**

From the agent's connections, choose PayPal and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[PayPal developer docs](https://developer.paypal.com)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Pendo
Source: https://getskydive.com/docs/integrations/pendo

Product analytics and in-app guides

## Example agent tasks

```text theme={null}
Connect Pendo and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Pendo.
```

```text theme={null}
Use the agent to interact with Pendo directly from your conversation.
```

## Connecting

Connect Pendo over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Pendo.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Pendo through the MCP server.

## What you share

Connecting grants the agent access to Pendo on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Permissions
Source: https://getskydive.com/docs/integrations/permissions

How agent access is scoped and credentials stay secure

Giving an agent real access to your tools is a big step. This page explains the security model behind every connection.

## The transparent proxy

All of an agent's outbound traffic flows through a **proxy** between the sandbox and the internet. Credentials are applied there. When the agent makes a request to a connected service, the proxy attaches the right token **on the wire**. The agent's own code and prompts never contain it.

```
Agent sandbox    Proxy    External service
                     
            attaches your credential here,
            never inside the agent
```

The result: the agent has the **capability** to act in a service without ever holding the **credential**.

## Why it's designed this way

Large language models shouldn't be trusted with raw secrets. A token in a prompt could surface in a log, a message, or a memory file. Skydive keeps credentials out of the agent entirely and injects them at the network layer. That removes the whole class of risk.

> **Secrets never leak**
>
> Tokens aren't in prompts, logs, or memory. They live outside the agent's reach.

> **Scoped per service**
>
> Each connection is limited to the service and permissions you granted.

> **You control the grant**
>
> Authorize narrowly or broadly, and revoke at any time.

> **Isolated sandboxes**
>
> Each agent runs in its own machine. One agent's access doesn't bleed into another's.

## Scoping access

You decide how much an agent can reach:

* **Per service**: connect only the tools a given agent needs.
* **Per resource**: for services like [GitHub](/docs/integrations/github), grant specific repositories rather than everything.
* **Per permission**: connections request only the scopes required for the work.

## Revoking access

Remove a connection or rotate a key at any time. The agent loses that access on its next request. No lingering tokens, no manual cleanup.

## A note on memory

[Memory](/docs/agents/memory) lives in the agent's [git repository](/docs/agents/repo), scoped to your workspace. It's shared knowledge for everyone with access to the agent. Treat it like a team wiki, not a vault. Keep genuine secrets in [secrets](/docs/integrations/secrets), not in things you tell the agent to remember.

> **[Connect a service securely](/docs/integrations/oauth)**
>
> See the authorization flow in practice.


# Pipedream
Source: https://getskydive.com/docs/integrations/pipedream

Let your agent run and orchestrate Pipedream workflows

Pipedream is a workflow automation platform built for developers. Connect it and your agent can trigger workflows, pass data between steps, and drive automation logic without you touching the Pipedream UI.

## What the agent can do

> **Trigger workflows**
>
> Start Pipedream workflows programmatically from a conversation or another agent step.

> **Pass data to steps**
>
> Send structured payloads into a workflow and capture the response.

> **Inspect runs**
>
> Check workflow execution history and surface errors or output.

> **Manage automations**
>
> Activate, pause, and reconfigure workflows through the agent.

## Example agent tasks

```text theme={null}
When a support ticket comes in, trigger the triage workflow and report what it returned.
```

```text theme={null}
Run the weekly report pipeline and send me the output when it finishes.
```

```text theme={null}
Check whether the nightly sync workflow ran successfully last night.
```

## What you share

Connecting gives the agent access to your Pipedream project workflows and execution context. See [Permissions](/docs/integrations/permissions).

## Connecting

**1. Add the Pipedream connection**

From the agent's connections, choose Pipedream. The platform connects you through Pipedream's hosted MCP server.

**2. Sign in to Pipedream**

A connect card appears. Sign in to your Pipedream account and approve access.

**3. The agent gets to work**

It can now trigger and inspect your Pipedream workflows.

## Learn more

> **[Pipedream documentation](https://pipedream.com/docs)**
>
> Guides, references, and the full Pipedream API.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Pipedrive
Source: https://getskydive.com/docs/integrations/pipedrive

Let your agent work with your Pipedrive CRM

Pipedrive is a sales CRM built around the pipeline. Connect it and your agent can read deals, people, and organizations, and keep them up to date.

## Two ways to connect

Pipedrive offers both an MCP connection and OAuth. MCP is the preferred path for most workflows.

* **Pipedrive MCP (recommended).** Broad CRM access through Pipedrive's hosted MCP server, with permission-aware tools. See [Connect with MCP](#connect-with-mcp).
* **Pipedrive OAuth.** The original sign-in integration. See [Connect with OAuth](#connect-with-oauth).

## What the agent can do

> **Read deals**
>
> Pull deals and where they sit in the pipeline.

> **Manage contacts**
>
> Work with people and organizations.

> **Update records**
>
> Change deal stage, value, or contact details.

> **Log activity**
>
> Record notes, calls, and follow-ups.

## Example agent tasks

```text theme={null}
When a deal moves to Proposal, draft the quote and attach it to the deal.
```

```text theme={null}
Each Monday, list deals that stalled last week and suggest a next step for each.
```

```text theme={null}
Log this call on the right deal and set a follow-up for Friday.
```

## Connect with MCP

**1. Add the Pipedrive connection**

From the agent's connections, choose Pipedrive. The platform connects you through Pipedrive's hosted MCP server.

**2. Sign in to Pipedrive**

A connect card appears. Sign in to your Pipedrive account and approve access. The MCP inherits your Pipedrive user permissions.

**3. The agent gets to work**

It can now read and update deals, contacts, and activity in your Pipedrive CRM.

## Connect with OAuth

## What you share

During the connect flow you authorize what the agent can reach. It works only within what you grant. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

**1. Authorize Pipedrive for the agent**

From the agent's connections, choose Pipedrive and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[Pipedrive API docs](https://developers.pipedrive.com/docs/api/v1)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Plain
Source: https://getskydive.com/docs/integrations/plain

Customer support platform for B2B teams

## Example agent tasks

```text theme={null}
Connect Plain and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Plain.
```

```text theme={null}
Use the agent to interact with Plain directly from your conversation.
```

## Connecting

Connect Plain over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Plain.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Plain through the MCP server.

## What you share

Connecting grants the agent access to Plain on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Polar MCP server (production)
Source: https://getskydive.com/docs/integrations/polar

Payments and billing infrastructure for developers

## Example agent tasks

```text theme={null}
Connect Polar MCP server (production) and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Polar MCP server (production).
```

```text theme={null}
Use the agent to interact with Polar MCP server (production) directly from your conversation.
```

## Connecting

Connect Polar (production) over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Polar MCP server (production).

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Polar MCP server (production) through the MCP server.

## What you share

Connecting grants the agent access to Polar MCP server (production) on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Portal
Source: https://getskydive.com/docs/integrations/portal

Your agent can act on your own machine, with your permission

Your agent runs in its own cloud sandbox, but some work lives on your laptop: a dev server on `localhost`, a command-line tool you have installed, files on your disk, a browser signed into your accounts. Portal lets your agent reach that machine when you run **Skydive Desktop** and grant that agent access. It needs the desktop app or the [Skydive CLI](/docs/cli/overview) running on that machine: you can't grant portal access from a phone or tablet browser, so work that touches your computer waits until you're back at one.

![Bodhi connecting to a laptop](https://mintcdn.com/skydive/2VR7bKqFZZqW4hyR/images/agents/portal-bodhi.png?fit=max\&auto=format\&n=2VR7bKqFZZqW4hyR\&q=85\&s=bad07f279197109e4dbe7b2b95e40dce)

## How it works

Portal is off until you turn it on, per agent. Nothing reaches your computer until you say yes.

**1. Install Skydive Desktop**

[Skydive Desktop](/docs/desktop/mac) is the small Mac app that connects your machine to your agents. When an agent first asks for access, its request card links the download. Not on a Mac? The [Skydive CLI](/docs/cli/overview) does the same job from a terminal.

**2. Approve the agent**

The agent asks for permission and a card appears in your chat. Approve it, and that one agent can reach your machine.

**3. Ask for the work**

Tell the agent what you need done on your computer. It runs the command, reaches the local server, or drives your browser, and reports back.

Access is granted per agent. Revoke it at any time from [Skydive Desktop](/docs/desktop/mac#letting-an-agent-use-your-mac).

## What it unlocks

> **Local servers**
>
> Reach a dev server or local app running on `localhost`.

> **Local tools**
>
> Run a command-line tool or script installed on your machine.

> **Your browser**
>
> Drive your own browser, signed into your accounts.

> **Local files**
>
> Read and work with files that live on your disk.

## Driving your browser

When a task needs your accounts (anything behind a login), the agent can drive your own signed-in browser instead of asking for credentials. It tells you before it takes control, keeps it brief, and hands control back when done.

For public pages that don't need your logins, the agent uses its own [browser](/docs/capabilities/browser) in the cloud. It never has to touch your machine for those.

## Good uses

* Restarting a local dev server and tailing its logs
* Running a build or a script you have set up locally
* Pulling a report from a tool that only runs on your machine
* Posting on your behalf in a web app you're already logged into

Ask for it in plain language:

```text theme={null}
My dev server on localhost:3000 is throwing errors. Pull the logs and
tell me what's breaking.
```

> [!NOTE]
>
> Portal is default-deny and granted per agent. An agent can only reach your computer after you approve it in Skydive Desktop. You can revoke that access at any time.


# Posthog
Source: https://getskydive.com/docs/integrations/posthog

Product analytics, feature flags and session replay

## Example agent tasks

```text theme={null}
Connect Posthog and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Posthog.
```

```text theme={null}
Use the agent to interact with Posthog directly from your conversation.
```

## Connecting

Connect Posthog over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Posthog.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Posthog through the MCP server.

## What you share

Connecting grants the agent access to Posthog on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Primitive
Source: https://getskydive.com/docs/integrations/primitive

Email infrastructure for AI agents

## What the agent can do

> **List domains**
>
> List all inbound domains for the organization, both verified and unverified.

> **Add domain**
>
> Claim a new domain and receive the DNS records to publish.

> **Verify domain**
>
> Check DNS records for a domain claim (MX, TXT, SPF, DKIM, DMARC).

> **Download domain zone file**
>
> Download a BIND-format DNS zone file for a domain.

## Example agent tasks

```text theme={null}
List everything available in Primitive and tell me what stands out.
```

```text theme={null}
Pull the latest data from Primitive and give me a summary.
```

## Connecting

Connect Primitive over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Primitive.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Primitive through the MCP server.

## What you share

Connecting grants the agent access to Primitive on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Printful
Source: https://getskydive.com/docs/integrations/printful

Let your agent work with your Printful orders and products

Printful handles print-on-demand fulfillment. Connect it and your agent can read your products and orders.

## What the agent can do

> **Read products**
>
> See your products and variants.

> **Read orders**
>
> Pull order status and details.

> **Track fulfillment**
>
> See what's printing and shipping.

> **See summaries**
>
> Roll up your activity.

## Example agent tasks

```text theme={null}
When an order ships, notify the customer with tracking automatically.
```

```text theme={null}
Every morning, flag any order stuck in production too long.
```

```text theme={null}
Tell me which products are selling and which to restock.
```

## What you share

During the connect flow you authorize what the agent can reach. It works only within what you grant. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Authorize Printful for the agent**

From the agent's connections, choose Printful and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[Printful API docs](https://developers.printful.com/docs/)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Prisma
Source: https://getskydive.com/docs/integrations/prisma

Database ORM and toolkit for developers

## Example agent tasks

```text theme={null}
Connect Prisma and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Prisma.
```

```text theme={null}
Use the agent to interact with Prisma directly from your conversation.
```

## Connecting

Connect Prisma over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Prisma.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Prisma through the MCP server.

## What you share

Connecting grants the agent access to Prisma on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Prismic
Source: https://getskydive.com/docs/integrations/prismic

Headless CMS with reusable page-section slices

## Example agent tasks

```text theme={null}
Connect Prismic and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Prismic.
```

```text theme={null}
Use the agent to interact with Prismic directly from your conversation.
```

## Connecting

Connect Prismic Asset with an API key.

**1. Add the connection**

Generate an API key from your Prismic account settings.

**2. Authorize**

Copy the key. It is only shown once.

**3. The agent gets to work**

Paste it when prompted in the agent's connection card for Prismic.

## What you share

Connecting grants the agent access to Prismic on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Probely
Source: https://getskydive.com/docs/integrations/probely

Web application vulnerability scanning

## Example agent tasks

```text theme={null}
Connect Probely and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Probely.
```

```text theme={null}
Use the agent to interact with Probely directly from your conversation.
```

## Connecting

Connect Probely with an API key.

**1. Add the connection**

Generate an API key from your Probely account settings.

**2. Authorize**

Copy the key. It is only shown once.

**3. The agent gets to work**

Paste it when prompted in the agent's connection card for Probely.

## What you share

Connecting grants the agent access to Probely on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Productboard
Source: https://getskydive.com/docs/integrations/productboard

Let your agent read your Productboard data

Productboard is a product management tool. Connect it and your agent can read product specs, search features, and keep your Productboard up to date.

## Two ways to connect

Productboard now runs a hosted MCP server. Connect it and your agent can pull and update spec content directly, without copy-pasting.

* **Productboard MCP (recommended).** Connect Productboard's MCP server once and the agent can read and update product specs, search features, post comments on specs. See [Connect with MCP](#connect-with-mcp).
* **Productboard OAuth.** The original sign-in integration, scoped to what you grant. See [Connect with OAuth](#connect-with-oauth).

## What the agent can do

> **Read specs**
>
> Pull the current content of any spec your account can reach.

> **Search features**
>
> Find features and initiatives by name, owner, or tag.

> **Update specs**
>
> Propose and apply edits to spec content.

> **Post comments**
>
> Add questions or updates directly on a spec.

## Example agent tasks

```text theme={null}
Show me the spec for the notifications feature.
```

```text theme={null}
Update the status of the comments API initiative to In Progress.
```

```text theme={null}
Summarize the open features tagged Q3 and their current owners.
```

## Connect with MCP

**1. Add the Productboard connection**

From the agent's connections, choose Productboard. The platform connects you through Productboard's MCP server.

**2. Sign in to Productboard**

A connect card appears. Sign in to your Productboard account and approve.

**3. The agent gets to work**

It can now read and update product specs, search features, post comments on specs through the MCP server.

> [!NOTE]
>
> MCP is the first-class connection when both options are available. See [Tools & MCP](/docs/capabilities/tools-and-mcp).

## Connect with OAuth

## What you share

During the connect flow you authorize what the agent can reach. It works only within what you grant. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

**1. Authorize Productboard for the agent**

From the agent's connections, choose Productboard and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[Productboard developer docs](https://developer.productboard.com)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# PlanetScale MCP server
Source: https://getskydive.com/docs/integrations/pscale

Serverless MySQL database platform

## Example agent tasks

```text theme={null}
Connect PlanetScale MCP server and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from PlanetScale MCP server.
```

```text theme={null}
Use the agent to interact with PlanetScale MCP server directly from your conversation.
```

## Connecting

Connect PlanetScale over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add PlanetScale.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach PlanetScale through the MCP server.

## What you share

Connecting grants the agent access to PlanetScale MCP server on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Pulumi
Source: https://getskydive.com/docs/integrations/pulumi

Infrastructure as code for cloud deployment

## Example agent tasks

```text theme={null}
Connect Pulumi and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Pulumi.
```

```text theme={null}
Use the agent to interact with Pulumi directly from your conversation.
```

## Connecting

Connect Pulumi over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Pulumi.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Pulumi through the MCP server.

## What you share

Connecting grants the agent access to Pulumi on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Qovery
Source: https://getskydive.com/docs/integrations/qovery

Cloud deployment and DevOps automation

## Example agent tasks

```text theme={null}
Connect Qovery and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Qovery.
```

```text theme={null}
Use the agent to interact with Qovery directly from your conversation.
```

## Connecting

**MCP server**

Connect Qovery over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Qovery.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Qovery through the MCP server.

**API key**

Connect Qovery with an API key.

**1. Add the connection**

Generate an API key from your Qovery account settings.

**2. Authorize**

Copy the key. It is only shown once.

**3. The agent gets to work**

Paste it when prompted in the agent's connection card for Qovery.

## What you share

Connecting grants the agent access to Qovery on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Quicknode
Source: https://getskydive.com/docs/integrations/quicknode

Blockchain node infrastructure and APIs

## Example agent tasks

```text theme={null}
Connect Quicknode and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Quicknode.
```

```text theme={null}
Use the agent to interact with Quicknode directly from your conversation.
```

## Connecting

Connect Quicknode over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Quicknode.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Quicknode through the MCP server.

## What you share

Connecting grants the agent access to Quicknode on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Quo
Source: https://getskydive.com/docs/integrations/quo

AI-powered business phone system

## Example agent tasks

```text theme={null}
Connect Quo and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Quo.
```

```text theme={null}
Use the agent to interact with Quo directly from your conversation.
```

## Connecting

Connect Quo over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Quo.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Quo through the MCP server.

## What you share

Connecting grants the agent access to Quo on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Ramp
Source: https://getskydive.com/docs/integrations/ramp

Let your agent manage corporate cards, transactions, and spend in Ramp

Ramp is a corporate card and spend management platform. Connect it and your agent can review transactions, manage cards, track budgets, and surface spending insights across your company.

## What the agent can do

> **Review transactions**
>
> Retrieve recent charges, filter by merchant or category, and surface unusual spend.

> **Manage cards**
>
> Look up virtual and physical cards, check limits, and take actions on active cards.

> **Track budgets**
>
> Check budget utilization by department or team and flag overages as they happen.

> **Surface spend insights**
>
> Summarize company spending by category, vendor, or time period.

## Example agent tasks

```text theme={null}
Show me all transactions over $500 from last month, grouped by category.
```

```text theme={null}
Which team has the highest spend this quarter, and what are the top vendors?
```

```text theme={null}
List all active virtual cards and their current monthly limits.
```

## What you share

During the connect flow you sign in to your Ramp account. The agent works within the permissions your Ramp role allows. The token is stored securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Add the Ramp connection**

From the agent's connections, choose Ramp. The platform connects through Ramp's MCP server.

**2. Sign in to Ramp**

A connect card appears. Sign in to your Ramp account and approve access.

**3. The agent gets to work**

It can now review transactions, manage cards, and track spend across your organization.

## Learn more

> **[Ramp docs](https://docs.ramp.com)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Raptive Food MCP
Source: https://getskydive.com/docs/integrations/raptive

Ad management for food content creators

## Example agent tasks

```text theme={null}
Connect Raptive Food MCP and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Raptive Food MCP.
```

```text theme={null}
Use the agent to interact with Raptive Food MCP directly from your conversation.
```

## Connecting

Connect Raptive Food over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Raptive Food.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Raptive Food through the MCP server.

## What you share

Connecting grants the agent access to Raptive Food MCP on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Razorpay
Source: https://getskydive.com/docs/integrations/razorpay

Indian payments and banking platform

## Example agent tasks

```text theme={null}
Connect Razorpay and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Razorpay.
```

```text theme={null}
Use the agent to interact with Razorpay directly from your conversation.
```

## Connecting

Connect Razorpay over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Razorpay.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Razorpay through the MCP server.

## What you share

Connecting grants the agent access to Razorpay on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Read AI
Source: https://getskydive.com/docs/integrations/read-ai

AI meeting notes, summaries and transcripts

Read AI captures meeting notes, summaries, and transcripts automatically. Connect it and your agent can pull what was said, extract action items, and surface insights from your recorded meetings without you digging through recordings yourself.

## What the agent can do

> **Pull meeting summaries**
>
> Get concise summaries of past meetings, including key decisions and topics covered.

> **Retrieve transcripts**
>
> Access full transcripts from recorded meetings for detailed review or quoting.

> **Surface action items**
>
> Extract follow-ups and tasks assigned during meetings so nothing falls through the cracks.

> **Search across meetings**
>
> Find what was said about a topic or by a person across your recorded meeting history.

## Example agent tasks

```text theme={null}
Summarize yesterday's product sync and list any action items assigned to me.
```

```text theme={null}
What did we decide about the Q4 roadmap in last week's planning meeting?
```

```text theme={null}
Pull the transcript from the all-hands on August 15th.
```

## What you share

Connecting gives the agent access to your Read AI account, including meeting summaries, transcripts, and action items. See [Permissions](/docs/integrations/permissions).

## Connecting

**1. Add the Read AI connection**

From the agent's connections, choose Read AI. The platform connects you through Read AI's MCP server.

**2. Sign in to Read AI**

A connect card appears. Sign in to your Read AI account and approve access.

**3. The agent gets to work**

It can now pull summaries, transcripts, and action items from your recorded meetings.

## Learn more

> **[Read AI](https://read.ai)**
>
> Learn more about Read AI and how it captures meetings.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Readwise
Source: https://getskydive.com/docs/integrations/readwise

Reading highlights and spaced repetition

## Example agent tasks

```text theme={null}
Connect Readwise and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Readwise.
```

```text theme={null}
Use the agent to interact with Readwise directly from your conversation.
```

## Connecting

Connect Readwise over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Readwise.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Readwise through the MCP server.

## What you share

Connecting grants the agent access to Readwise on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Rebilly
Source: https://getskydive.com/docs/integrations/rebilly

Subscription billing and payment orchestration

## Example agent tasks

```text theme={null}
Connect Rebilly and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Rebilly.
```

```text theme={null}
Use the agent to interact with Rebilly directly from your conversation.
```

## Connecting

Connect Rebilly with an API key.

**1. Add the connection**

Generate an API key from your Rebilly account settings.

**2. Authorize**

Copy the key. It is only shown once.

**3. The agent gets to work**

Paste it when prompted in the agent's connection card for Rebilly.

## What you share

Connecting grants the agent access to Rebilly on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# ReCommerce MCP Server
Source: https://getskydive.com/docs/integrations/recommerce

European refurbished device buyback and resale

## Example agent tasks

```text theme={null}
Connect ReCommerce MCP Server and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from ReCommerce MCP Server.
```

```text theme={null}
Use the agent to interact with ReCommerce MCP Server directly from your conversation.
```

## Connecting

Connect Recommerce Trade-in over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add ReCommerce.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach ReCommerce through the MCP server.

## What you share

Connecting grants the agent access to ReCommerce MCP Server on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Redocly MCP Server
Source: https://getskydive.com/docs/integrations/redocly

OpenAPI documentation tooling

## Example agent tasks

```text theme={null}
Connect Redocly MCP Server and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Redocly MCP Server.
```

```text theme={null}
Use the agent to interact with Redocly MCP Server directly from your conversation.
```

## Connecting

Connect Redocly over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Redocly.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Redocly through the MCP server.

## What you share

Connecting grants the agent access to Redocly MCP Server on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Reducto
Source: https://getskydive.com/docs/integrations/reducto

Document parsing and data extraction for AI

## Example agent tasks

```text theme={null}
Connect Reducto and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Reducto.
```

```text theme={null}
Use the agent to interact with Reducto directly from your conversation.
```

## Connecting

Connect Reducto Docs over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Reducto.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Reducto through the MCP server.

## What you share

Connecting grants the agent access to Reducto on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Refiner
Source: https://getskydive.com/docs/integrations/refiner

In-app surveys and customer feedback

## Example agent tasks

```text theme={null}
Connect Refiner and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Refiner.
```

```text theme={null}
Use the agent to interact with Refiner directly from your conversation.
```

## Connecting

Connect Refiner over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Refiner.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Refiner through the MCP server.

## What you share

Connecting grants the agent access to Refiner on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Releases MCP server
Source: https://getskydive.com/docs/integrations/releases

Unified registry of product changelogs

## Example agent tasks

```text theme={null}
Connect Releases MCP server and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Releases MCP server.
```

```text theme={null}
Use the agent to interact with Releases MCP server directly from your conversation.
```

## Connecting

Connect Releases over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Releases.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Releases through the MCP server.

## What you share

Connecting grants the agent access to Releases MCP server on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Background Removal
Source: https://getskydive.com/docs/integrations/remove

Image background removal API

## Example agent tasks

```text theme={null}
Connect Background Removal and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Background Removal.
```

```text theme={null}
Use the agent to interact with Background Removal directly from your conversation.
```

## Connecting

Connect Background Removal with an API key.

**1. Add the connection**

Generate an API key from your Background Removal account settings.

**2. Authorize**

Copy the key. It is only shown once.

**3. The agent gets to work**

Paste it when prompted in the agent's connection card for Background Removal.

## What you share

Connecting grants the agent access to Background Removal on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Replicate
Source: https://getskydive.com/docs/integrations/replicate

Hosted AI model inference API

## Example agent tasks

```text theme={null}
Connect Replicate and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Replicate.
```

```text theme={null}
Use the agent to interact with Replicate directly from your conversation.
```

## Connecting

**MCP server**

Connect Replicate over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Replicate.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Replicate through the MCP server.

**API key**

Connect Replicate with an API key.

**1. Add the connection**

Generate an API key from your Replicate account settings.

**2. Authorize**

Copy the key. It is only shown once.

**3. The agent gets to work**

Paste it when prompted in the agent's connection card for Replicate.

## What you share

Connecting grants the agent access to Replicate on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Replit
Source: https://getskydive.com/docs/integrations/replit-mcp

Cloud-based coding and app deployment

## Example agent tasks

```text theme={null}
Connect Replit and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Replit.
```

```text theme={null}
Use the agent to interact with Replit directly from your conversation.
```

## Connecting

Connect Replit over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Replit.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Replit through the MCP server.

## What you share

Connecting grants the agent access to Replit on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Resend
Source: https://getskydive.com/docs/integrations/resend

Transactional email API for developers

## Example agent tasks

```text theme={null}
Connect Resend and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Resend.
```

```text theme={null}
Use the agent to interact with Resend directly from your conversation.
```

## Connecting

Connect Resend over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Resend.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Resend through the MCP server.

## What you share

Connecting grants the agent access to Resend on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Spectra Assure Community
Source: https://getskydive.com/docs/integrations/reversinglabs

Software supply chain security scanning

## Example agent tasks

```text theme={null}
Connect Spectra Assure Community and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Spectra Assure Community.
```

```text theme={null}
Use the agent to interact with Spectra Assure Community directly from your conversation.
```

## Connecting

Connect Spectra Assure Community with an API key.

**1. Add the connection**

Generate an API key from your Spectra Assure Community account settings.

**2. Authorize**

Copy the key. It is only shown once.

**3. The agent gets to work**

Paste it when prompted in the agent's connection card for Spectra Assure Community.

## What you share

Connecting grants the agent access to Spectra Assure Community on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Roam HQ
Source: https://getskydive.com/docs/integrations/roam

Let your agent work in your Roam HQ workspace

Roam HQ is a workspace that combines meetings, chat, calendar, and team knowledge in one place. Connect it and your agent can join the workspace, surface conversation context, and help your team act on what's happening.

## What the agent can do

> **Access workspace context**
>
> Pull messages, meeting notes, and team knowledge from your Roam workspace.

> **Surface discussions**
>
> Find relevant conversations and surface context from past meetings or threads.

> **Work with calendar**
>
> Check scheduling, meeting status, and what's coming up for your team.

> **Read workspace knowledge**
>
> Pull documented processes, decisions, and shared resources from the workspace.

## Example agent tasks

```text theme={null}
What did the team discuss in this week's standups? Summarize the key decisions.
```

```text theme={null}
Find the notes from last Thursday's product planning meeting.
```

```text theme={null}
What's on my calendar tomorrow and what context do I need for each meeting?
```

## What you share

Connecting gives the agent access to your Roam HQ workspace, including messages, meeting content, and calendar data within your permissions. Workspace admins may need to approve the connection. See [Permissions](/docs/integrations/permissions).

## Connecting

**1. Add the Roam HQ connection**

From the agent's connections, choose Roam HQ. The platform connects you through Roam's hosted MCP server.

**2. Sign in to Roam**

A connect card appears. Sign in to your Roam account and approve access. Workspace admins may need to approve first.

**3. The agent gets to work**

It can now surface meeting context, workspace knowledge, and calendar information from Roam.

## Learn more

> **[Roam HQ support](https://ro.am/support)**
>
> Product guides and support resources.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Rootly
Source: https://getskydive.com/docs/integrations/rootly

Incident response with AI root-cause confidence scoring

## Example agent tasks

```text theme={null}
Connect Rootly and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Rootly.
```

```text theme={null}
Use the agent to interact with Rootly directly from your conversation.
```

## Connecting

Connect Rootly over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Rootly.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Rootly through the MCP server.

## What you share

Connecting grants the agent access to Rootly on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Sakari Messaging
Source: https://getskydive.com/docs/integrations/sakari

SMS marketing platform

## Example agent tasks

```text theme={null}
Connect Sakari Messaging and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Sakari Messaging.
```

```text theme={null}
Use the agent to interact with Sakari Messaging directly from your conversation.
```

## Connecting

Connect Sakari Messaging with an API key.

**1. Add the connection**

Generate an API key from your Sakari Messaging account settings.

**2. Authorize**

Copy the key. It is only shown once.

**3. The agent gets to work**

Paste it when prompted in the agent's connection card for Sakari Messaging.

## What you share

Connecting grants the agent access to Sakari Messaging on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Sanity
Source: https://getskydive.com/docs/integrations/sanity

Headless CMS for content operations

## Example agent tasks

```text theme={null}
Connect Sanity and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Sanity.
```

```text theme={null}
Use the agent to interact with Sanity directly from your conversation.
```

## Connecting

Connect Sanity over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Sanity.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Sanity through the MCP server.

## What you share

Connecting grants the agent access to Sanity on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# ScrapingBee Remote MCP Server
Source: https://getskydive.com/docs/integrations/scrapingbee

Web scraping and browser automation API

## Example agent tasks

```text theme={null}
Connect ScrapingBee Remote MCP Server and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from ScrapingBee Remote MCP Server.
```

```text theme={null}
Use the agent to interact with ScrapingBee Remote MCP Server directly from your conversation.
```

## Connecting

Connect ScrapingBee Remote over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add ScrapingBee Remote.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach ScrapingBee Remote through the MCP server.

## What you share

Connecting grants the agent access to ScrapingBee Remote MCP Server on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Secrets
Source: https://getskydive.com/docs/integrations/secrets

Store credentials your agent can use without ever seeing them

Secrets are how you give your agent an API key or token for a service that isn't connected through one-click OAuth. The agent uses it **without ever seeing the raw value**.

## Your agent never sees the secret

The model driving your agent is never trusted with raw credentials, which could otherwise end up in a log, a memory file, or a message. A secret is stored encrypted and lives **outside the agent's reach**. A proxy injects it into requests at the network layer. The agent gets the *capability* without holding the *credential*.

The same principle applies to every connection on Skydive. See [Permissioning](/docs/integrations/permissions) for the full security model.

## When to use a secret

* A service offers an **API key or token** but not OAuth
* You have an internal API the agent should call
* A tool or MCP server needs a token to authenticate

## How it works

You don't have to set secrets up in advance. When a task calls for one, the agent presents a **secure card** for you to paste the key into. You can also offer one anytime by telling the agent you want to give it a credential.

**1. Add the secret**

Paste the key into the secure card and name it, so the agent knows what it's for.

**2. It's stored, not shown**

The value is encrypted at rest. The agent can reference it by name and use it, but never reads the plaintext.

**3. Used on the wire**

When the agent makes a request to that service, a proxy injects the credential into the request. The request authenticates correctly while the secret stays out of prompts and logs.

## Managing secrets

Add, update, and remove secrets at any time. Rotate a key by updating its value. The agent picks up the new one on its next request.

An agent's **Manage secrets** panel supports bulk work. Add several keys at once: paste a whole `.env` block or import one from a file, and it splits into rows. Copy secrets out as `KEY=value` lines, or download the whole set as a `.env` file. Moving a group of credentials between agents is a single step.

> **[Prefer OAuth where you can](/docs/integrations/oauth)**
>
> One-click connections refresh themselves and are cleaner than manual keys.


# SecureLend AI Agents
Source: https://getskydive.com/docs/integrations/securelend

AI underwriting infrastructure for institutional lenders

## Example agent tasks

```text theme={null}
Connect SecureLend AI Agents and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from SecureLend AI Agents.
```

```text theme={null}
Use the agent to interact with SecureLend AI Agents directly from your conversation.
```

## Connecting

Connect SecureLend AI Agents over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add SecureLend AI Agents.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach SecureLend AI Agents through the MCP server.

## What you share

Connecting grants the agent access to SecureLend AI Agents on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Semaphore
Source: https://getskydive.com/docs/integrations/semaphoreci

Fast CI/CD pipelines with simple per-seat pricing

## Example agent tasks

```text theme={null}
Connect Semaphore and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Semaphore.
```

```text theme={null}
Use the agent to interact with Semaphore directly from your conversation.
```

## Connecting

Connect Semaphore over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Semaphore.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Semaphore through the MCP server.

## What you share

Connecting grants the agent access to Semaphore on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Send
Source: https://getskydive.com/docs/integrations/send

Publishes and hosts artifacts made in Claude

## Example agent tasks

```text theme={null}
Connect Send and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Send.
```

```text theme={null}
Use the agent to interact with Send directly from your conversation.
```

## Connecting

Connect Send over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Send.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Send through the MCP server.

## What you share

Connecting grants the agent access to Send on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Sentry
Source: https://getskydive.com/docs/integrations/sentry

Application error tracking and monitoring

## Example agent tasks

```text theme={null}
Connect Sentry and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Sentry.
```

```text theme={null}
Use the agent to interact with Sentry directly from your conversation.
```

## Connecting

Connect Sentry over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Sentry.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Sentry through the MCP server.

## What you share

Connecting grants the agent access to Sentry on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# SerpApi
Source: https://getskydive.com/docs/integrations/serpapi

Search engine results scraping API

## Example agent tasks

```text theme={null}
Connect SerpApi and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from SerpApi.
```

```text theme={null}
Use the agent to interact with SerpApi directly from your conversation.
```

## Connecting

**MCP server**

Connect SerpApi over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add SerpApi.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach SerpApi through the MCP server.

**API key**

Connect SerpApi Search with an API key.

**1. Add the connection**

Generate an API key from your SerpApi account settings.

**2. Authorize**

Copy the key. It is only shown once.

**3. The agent gets to work**

Paste it when prompted in the agent's connection card for SerpApi.

## What you share

Connecting grants the agent access to SerpApi on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Serpstat
Source: https://getskydive.com/docs/integrations/serpstat

SEO and keyword research platform

## Example agent tasks

```text theme={null}
Connect Serpstat and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Serpstat.
```

```text theme={null}
Use the agent to interact with Serpstat directly from your conversation.
```

## Connecting

Connect Serpstat over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Serpstat.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Serpstat through the MCP server.

## What you share

Connecting grants the agent access to Serpstat on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# ShareOut
Source: https://getskydive.com/docs/integrations/shareout

Agentic publishing for pages and reports

## Example agent tasks

```text theme={null}
Connect ShareOut and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from ShareOut.
```

```text theme={null}
Use the agent to interact with ShareOut directly from your conversation.
```

## Connecting

Connect ShareOut with an API key.

**1. Add the connection**

Generate an API key from your ShareOut account settings.

**2. Authorize**

Copy the key. It is only shown once.

**3. The agent gets to work**

Paste it when prompted in the agent's connection card for ShareOut.

## What you share

Connecting grants the agent access to ShareOut on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Shop MCP server
Source: https://getskydive.com/docs/integrations/shop

Shopify's consumer shopping app

## Example agent tasks

```text theme={null}
Connect Shop MCP server and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Shop MCP server.
```

```text theme={null}
Use the agent to interact with Shop MCP server directly from your conversation.
```

## Connecting

Connect Shop over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Shop.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Shop through the MCP server.

## What you share

Connecting grants the agent access to Shop MCP server on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Shortcut
Source: https://getskydive.com/docs/integrations/shortcut

Project management for software teams

## Example agent tasks

```text theme={null}
Connect Shortcut and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Shortcut.
```

```text theme={null}
Use the agent to interact with Shortcut directly from your conversation.
```

## Connecting

Connect Shortcut over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Shortcut.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Shortcut through the MCP server.

## What you share

Connecting grants the agent access to Shortcut on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# SIGNL4
Source: https://getskydive.com/docs/integrations/signl4

Mobile alerting and on-call escalation

## Example agent tasks

```text theme={null}
Connect SIGNL4 and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from SIGNL4.
```

```text theme={null}
Use the agent to interact with SIGNL4 directly from your conversation.
```

## Connecting

Connect SIGNL4 with an API key.

**1. Add the connection**

Generate an API key from your SIGNL4 account settings.

**2. Authorize**

Copy the key. It is only shown once.

**3. The agent gets to work**

Paste it when prompted in the agent's connection card for SIGNL4.

## What you share

Connecting grants the agent access to SIGNL4 on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# SignNow
Source: https://getskydive.com/docs/integrations/signnow

Let your agent work with your SignNow documents

SignNow handles e-signatures. Connect it and your agent can read documents and signature status and help send documents for signing.

## What the agent can do

> **Read documents**
>
> See documents and who has signed.

> **Check status**
>
> Track what's pending or complete.

> **Send for signature**
>
> Start a signing request from a document.

> **Find a document**
>
> Locate a specific document.

## Example agent tasks

```text theme={null}
Which SignNow documents are still awaiting signature?
```

```text theme={null}
What's the status of the vendor agreement?
```

```text theme={null}
Send this contract to the client for signature.
```

## What you share

During the connect flow you authorize what the agent can reach. It works only within what you grant. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Authorize SignNow for the agent**

From the agent's connections, choose SignNow and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[SignNow API docs](https://docs.signnow.com/)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# SigNoz
Source: https://getskydive.com/docs/integrations/signoz

Open source observability and APM platform

## Example agent tasks

```text theme={null}
Connect SigNoz and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from SigNoz.
```

```text theme={null}
Use the agent to interact with SigNoz directly from your conversation.
```

## Connecting

Connect SigNoz over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add SigNoz.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach SigNoz through the MCP server.

## What you share

Connecting grants the agent access to SigNoz on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Skydive
Source: https://getskydive.com/docs/integrations/skydive

Let your agent build and manage agents in your Skydive workspace

Skydive is an agent-hosting platform. Connect it with an API key and your agent can list the agents in your workspace, create new ones, and manage API keys and secrets across them.

## What the agent can do

> **List and inspect agents**
>
> Fetch the agents in your workspace, check their status, and read their configuration.

> **Create agents**
>
> Provision new agents with a name, model, and description.

> **Manage API keys**
>
> Create, list, and rotate API keys for agents in your workspace.

> **Manage secrets**
>
> Write and update secrets for an agent so it can reach external services.

## Example agent tasks

```text theme={null}
Show me all the agents in my Skydive workspace and their current status.
```

```text theme={null}
Create a new agent called "Support bot" using Claude Sonnet.
```

```text theme={null}
Rotate the API key for the billing agent and update its STRIPE_KEY secret.
```

## What you share

Connecting grants the agent access to your Skydive workspace through an account-level API key. An edit-scoped key lets it create agents and manage their keys and secrets. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Create a Skydive API key**

Open [Skydive Settings](https://skydive.com/settings/account) and create an account-level API key. Choose edit scope if you want the agent to create agents or manage their keys and secrets. Copy the key when it appears.

**2. Add the Skydive connection**

From the agent's connections, choose Skydive and paste the key when prompted.

**3. The agent gets to work**

It can now list your workspace agents, create new ones, and manage their credentials through the Skydive API.

## Learn more

> **[Skydive docs](https://www.getskydive.com/docs)**
>
> Documentation for the Skydive platform.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# SkyWatch
Source: https://getskydive.com/docs/integrations/skywatch

Search satellite imagery, compare pricing, and explore Earth observation data for your location and time range

SkyWatch's MCP server gives your agent direct access to the EarthCache imagery marketplace: search for available scenes, compare prices across providers, price a tasking order, and browse the full satellite catalog. No account required to start.

## What the agent can do

> **Search archive imagery**
>
> Find orderable satellite scenes for any location and date range, with per-scene pricing and cloud cover data.

> **Calculate pricing**
>
> Get cost estimates for archive mosaics, tasking orders, and resolution tiers across all providers.

> **Browse satellites**
>
> List sensors in the SkyWatch catalog with their resolution, type, and archive or tasking availability.

> **Explore offerings**
>
> Compare archive and tasking prices side by side, filtered by provider, product, or resolution tier.

## Example agent tasks

```text theme={null}
Find high-resolution satellite imagery of the Port of Rotterdam from the last 90 days and show me pricing.
```

```text theme={null}
Which satellite providers have coverage of this farmland area in Iowa? Give me a budget breakdown.
```

```text theme={null}
What are the cheapest archive imagery options for my area with less than 20% cloud cover?
```

## What you share

This connection requires no sign-in or API key to browse the catalog and retrieve pricing. The agent has read access to publicly available imagery metadata, satellite specs, and product pricing. No personal information is shared unless you place an order through the SkyWatch platform directly.

## Connecting

**1. Add the SkyWatch connection**

From the agent's tools, add SkyWatch. Authorize access when the connect card appears.

**2. The agent gets to work**

The agent can now search imagery, retrieve pricing, and browse the satellite catalog.

## Learn more

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Slack
Source: https://getskydive.com/docs/integrations/slack

Team messaging and collaboration

## Example agent tasks

```text theme={null}
Connect Slack and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Slack.
```

```text theme={null}
Use the agent to interact with Slack directly from your conversation.
```

## Connecting

Connect your Slack account so your agent can work in it.

**1. Add the connection**

From the agent's connections, choose Slack and sign in.

**2. Authorize**

Approve the access. You can change this later.

**3. The agent gets to work**

The agent gets to work through the secure proxy.

## What you share

Connecting grants the agent access to Slack on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# SmartBear MCP Server (Swagger remote)
Source: https://getskydive.com/docs/integrations/smartbear

API design and testing tools

## Example agent tasks

```text theme={null}
Connect SmartBear MCP Server (Swagger remote) and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from SmartBear MCP Server (Swagger remote).
```

```text theme={null}
Use the agent to interact with SmartBear MCP Server (Swagger remote) directly from your conversation.
```

## Connecting

Connect SmartBear (BugSnag remote) over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add SmartBear MCP Server (Swagger remote).

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach SmartBear MCP Server (Swagger remote) through the MCP server.

## What you share

Connecting grants the agent access to SmartBear MCP Server (Swagger remote) on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Smarty
Source: https://getskydive.com/docs/integrations/smarty

Address validation and verification API

## Example agent tasks

```text theme={null}
Connect Smarty and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Smarty.
```

```text theme={null}
Use the agent to interact with Smarty directly from your conversation.
```

## Connecting

Connect Smarty over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Smarty.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Smarty through the MCP server.

## What you share

Connecting grants the agent access to Smarty on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# sms77.io
Source: https://getskydive.com/docs/integrations/sms77

SMS and voice messaging API

## Example agent tasks

```text theme={null}
Connect sms77.io and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from sms77.io.
```

```text theme={null}
Use the agent to interact with sms77.io directly from your conversation.
```

## Connecting

Connect sms77.io with an API key.

**1. Add the connection**

Generate an API key from your sms77.io account settings.

**2. Authorize**

Copy the key. It is only shown once.

**3. The agent gets to work**

Paste it when prompted in the agent's connection card for sms77.io.

## What you share

Connecting grants the agent access to sms77.io on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# SMTP2GO MCP server
Source: https://getskydive.com/docs/integrations/smtp2go

Transactional email delivery service

## Example agent tasks

```text theme={null}
Connect SMTP2GO MCP server and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from SMTP2GO MCP server.
```

```text theme={null}
Use the agent to interact with SMTP2GO MCP server directly from your conversation.
```

## Connecting

Connect SMTP2GO over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add SMTP2GO.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach SMTP2GO through the MCP server.

## What you share

Connecting grants the agent access to SMTP2GO MCP server on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Snowflake
Source: https://getskydive.com/docs/integrations/snowflake

Cloud data warehouse and analytics platform

Snowflake is a cloud data warehouse and analytics platform. Connect it with a programmatic access token and your agents can run SELECT queries against your databases, list databases, schemas, and tables, and inspect your data model, all through the Snowflake SQL API and REST API v2.

## What your agents can do

> **Run SQL queries**
>
> Execute SELECT statements through the SQL API and get results back, from quick lookups to analytical joins across tables.

> **Explore your data model**
>
> List databases, schemas, and tables so agents can map your warehouse before writing a query.

> **Build reports from live data**
>
> Pull numbers straight from Snowflake into summaries, dashboards, or answers without exporting CSVs.

> **Combine with your other tools**
>
> Join warehouse data with what agents already know from your other connections for richer analysis.

## Example agent tasks

```text theme={null}
Query our Snowflake warehouse for monthly active users by region over the last 6 months and summarize the trend.
```

```text theme={null}
List the databases and schemas in our account and describe what each staging table appears to contain.
```

```text theme={null}
Pull this week's revenue by product line from Snowflake and draft a short update for the team.
```

## What you share

Connecting grants your agents access to your Snowflake account through the SQL API and REST API v2 using a programmatic access token. Reads are the default; agents confirm with you before running any INSERT, UPDATE, DELETE, or DDL. The token reaches the Snowflake API only. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Generate a programmatic access token**

In Snowsight, go to My Profile > Programmatic access tokens > Generate new token. The token is shown once; copy it. Alternatively run `ALTER USER <user> ADD PROGRAMMATIC ACCESS TOKEN <name>` and copy the returned token. Also have your account URL ready (it looks like `https://<orgname>-<account>.snowflakecomputing.com`).

**2. Add the Snowflake connection**

From your agent's connections, choose Snowflake. Paste your token and account URL when prompted. The platform injects the token as the `Authorization: Bearer` header automatically.

**3. Your agents get to work**

They can now query and explore your warehouse through the SQL API and REST API v2.

> [!NOTE]
>
> If authentication fails with error code 390432 (network policy required, common on trial accounts), allow the token to bypass the network policy: in Snowsight, go to My Profile > Programmatic access tokens > token settings > Bypass requirement for network policy (valid up to 24 hours), or run `ALTER USER MODIFY PAT <name> SET MINS_TO_BYPASS_NETWORK_POLICY_REQUIREMENT = 1440`. If the token was rotated or revoked, generate a new one and update the connection.

## Learn more

> **[Snowflake SQL API](https://docs.snowflake.com/en/developer-guide/sql-api)**
>
> Reference for the SQL API your agents use to run queries.

> **[Programmatic access tokens](https://docs.snowflake.com/en/user-guide/programmatic-access-tokens)**
>
> How to generate and manage the tokens this connection uses.


# Spinify
Source: https://getskydive.com/docs/integrations/spinify

Sales gamification and leaderboards

## Example agent tasks

```text theme={null}
Connect Spinify and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Spinify.
```

```text theme={null}
Use the agent to interact with Spinify directly from your conversation.
```

## Connecting

Connect Spinify over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Spinify.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Spinify through the MCP server.

## What you share

Connecting grants the agent access to Spinify on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Splice
Source: https://getskydive.com/docs/integrations/splice

Music sample and loop library for producers

## Example agent tasks

```text theme={null}
Connect Splice and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Splice.
```

```text theme={null}
Use the agent to interact with Splice directly from your conversation.
```

## Connecting

Connect Splice over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Splice.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Splice through the MCP server.

## What you share

Connecting grants the agent access to Splice on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Splitwise
Source: https://getskydive.com/docs/integrations/splitwise

Let your agent work with your Splitwise expenses

Splitwise splits shared expenses between people. Connect it and your agent can read your groups, expenses, and balances.

## What the agent can do

> **Read expenses**
>
> See shared expenses.

> **See groups**
>
> Read your groups and members.

> **Check balances**
>
> See who owes what.

> **Add expenses**
>
> Record a new shared expense.

## Example agent tasks

```text theme={null}
After a trip, tally who owes what and send everyone their share.
```

```text theme={null}
When I add an expense, split it with the right group automatically.
```

```text theme={null}
Each month, tell me my balance across all my groups.
```

## What you share

During the connect flow you authorize what the agent can reach. It works only within what you grant. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Authorize Splitwise for the agent**

From the agent's connections, choose Splitwise and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[Splitwise API docs](https://dev.splitwise.com/)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Spotify
Source: https://getskydive.com/docs/integrations/spotify

Let your agent work with your Spotify music

Spotify is music and podcasts. Connect it and your agent can read your playlists, library, and listening, and help manage playlists.

## What the agent can do

> **Read library**
>
> See your saved tracks and albums.

> **Read playlists**
>
> Pull your playlists and their tracks.

> **Build playlists**
>
> Create playlists and add tracks.

> **Search**
>
> Find tracks and artists.

## Example agent tasks

```text theme={null}
Build me a focus playlist for deep work and refresh it weekly so it stays fresh.
```

```text theme={null}
Make a playlist that matches the vibe of this event.
```

```text theme={null}
Turn my recent favorites into a playlist I can share.
```

## What you share

During the connect flow you authorize what the agent can reach. It works only within what you grant. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Authorize Spotify for the agent**

From the agent's connections, choose Spotify and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[Spotify Web API docs](https://developer.spotify.com/documentation/web-api)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Spreedly Docs MCP Server
Source: https://getskydive.com/docs/integrations/spreedly

Payment method tokenization and orchestration

## Example agent tasks

```text theme={null}
Connect Spreedly Docs MCP Server and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Spreedly Docs MCP Server.
```

```text theme={null}
Use the agent to interact with Spreedly Docs MCP Server directly from your conversation.
```

## Connecting

Connect Spreedly Docs over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Spreedly Docs.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Spreedly Docs through the MCP server.

## What you share

Connecting grants the agent access to Spreedly Docs MCP Server on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Square
Source: https://getskydive.com/docs/integrations/square

Let your agent work with your Square sales and payments

Square runs payments, point of sale, and orders. Connect it and your agent can read sales, payments, and catalog items.

## What the agent can do

> **Read sales**
>
> See payments and orders.

> **Read catalog**
>
> Pull your items and prices.

> **Read customers**
>
> See customer details.

> **See summaries**
>
> Roll up your sales.

## Example agent tasks

```text theme={null}
Every evening, send me the day's sales and the top-selling items.
```

```text theme={null}
When sales dip below a normal day, flag it and tell me what's different.
```

```text theme={null}
Each Monday, tell me which products are trending up and which are slowing.
```

## What you share

During the connect flow you authorize what the agent can reach. It works only within what you grant. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Authorize Square for the agent**

From the agent's connections, choose Square and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[Square API docs](https://developer.squareup.com/reference/square)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Squareup
Source: https://getskydive.com/docs/integrations/squareup

Payments and point-of-sale

## What the agent can do

> **Search-brand-by-name**
>
> Given a restaurant brand name, return its locations.

> **Discover-nearby-restaurants**
>
> Discover restaurants near a given location.

> **Create-cart**
>
> Create a cart for a restaurant location using unit token and a list of item selections.

> **Get-single-menu**
>
> Get menu items for a restaurant location by brand name and address.

## Example agent tasks

```text theme={null}
Search Squareup for relevant results and summarize what you find.
```

```text theme={null}
Create a new entry in Squareup based on my notes.
```

```text theme={null}
Pull the latest data from Squareup and give me a summary.
```

## Connecting

**Cash App MCP**

Connect Cash App over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Squareup.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Squareup through the MCP server.

**Sign in with Squareup**

Connect your Squareup account so your agent can work in it.

**1. Add the connection**

From the agent's connections, choose Squareup and sign in.

**2. Authorize**

Approve the access. You can change this later.

**3. The agent gets to work**

The agent gets to work through the secure proxy.

## What you share

Connecting grants the agent access to Squareup on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Stable Baseline
Source: https://getskydive.com/docs/integrations/stablebaseline

Agent-managed company documentation and project workspace

## What the agent can do

> **Search tools**
>
> Search available tools by keyword or category.

> **List diagram types**
>
> List supported diagram types.

> **Get diagram type guide**
>
> Get DSL writing instructions for a diagram type.

> **List architecture icons**
>
> SOFTWARE & CLOUD architecture icons ONLY (AWS, Azure, GCP, Docker, Kubernetes, databases, message queues, dev tools, etc

## Example agent tasks

```text theme={null}
Search Stable Baseline for relevant results and summarize what you find.
```

```text theme={null}
List everything available in Stable Baseline and tell me what stands out.
```

```text theme={null}
Pull the latest data from Stable Baseline and give me a summary.
```

## Connecting

Connect Stable Baseline over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Stable Baseline.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Stable Baseline through the MCP server.

## What you share

Connecting grants the agent access to Stable Baseline on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Stack Exchange
Source: https://getskydive.com/docs/integrations/stackexchange

Let your agent search Stack Exchange and Stack Overflow

Stack Exchange powers Stack Overflow and dozens of Q\&A sites. Connect it and your agent can search questions and pull answers.

## What the agent can do

> **Search questions**
>
> Find questions on a topic.

> **Read answers**
>
> Pull the top answers to a question.

> **Your activity**
>
> See your questions and answers.

> **Browse tags**
>
> Find content by tag.

## Example agent tasks

```text theme={null}
When I hit this error, find the top-voted fix and walk me through it.
```

```text theme={null}
Research the current best practice for this and summarize the consensus.
```

```text theme={null}
Pull the three most useful answers on this topic into a short brief.
```

## What you share

During the connect flow you authorize what the agent can reach. It works only within what you grant. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Authorize Stack Exchange for the agent**

From the agent's connections, choose Stack Exchange and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[Stack Exchange API docs](https://api.stackexchange.com/docs)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Statsig
Source: https://getskydive.com/docs/integrations/statsig

Feature flags and experimentation platform

## Example agent tasks

```text theme={null}
Connect Statsig and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Statsig.
```

```text theme={null}
Use the agent to interact with Statsig directly from your conversation.
```

## Connecting

Connect Statsig over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Statsig.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Statsig through the MCP server.

## What you share

Connecting grants the agent access to Statsig on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Statuspage
Source: https://getskydive.com/docs/integrations/statuspage

Public status pages for incident communication

## Example agent tasks

```text theme={null}
Connect Statuspage and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Statuspage.
```

```text theme={null}
Use the agent to interact with Statuspage directly from your conversation.
```

## Connecting

Connect Statuspage with an API key.

**1. Add the connection**

Generate an API key from your Statuspage account settings.

**2. Authorize**

Copy the key. It is only shown once.

**3. The agent gets to work**

Paste it when prompted in the agent's connection card for Statuspage.

## What you share

Connecting grants the agent access to Statuspage on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Storyblok
Source: https://getskydive.com/docs/integrations/storyblok

Headless CMS with a visual editor for marketers

## Example agent tasks

```text theme={null}
Connect Storyblok and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Storyblok.
```

```text theme={null}
Use the agent to interact with Storyblok directly from your conversation.
```

## Connecting

**MCP server**

Connect Storyblok over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Storyblok.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Storyblok through the MCP server.

**API key**

Connect GraphQL with an API key.

**1. Add the connection**

Generate an API key from your Storyblok account settings.

**2. Authorize**

Copy the key. It is only shown once.

**3. The agent gets to work**

Paste it when prompted in the agent's connection card for Storyblok.

## What you share

Connecting grants the agent access to Storyblok on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Streak
Source: https://getskydive.com/docs/integrations/streak

CRM built inside Gmail

## Example agent tasks

```text theme={null}
Connect Streak and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Streak.
```

```text theme={null}
Use the agent to interact with Streak directly from your conversation.
```

## Connecting

Connect Streak over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Streak.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Streak through the MCP server.

## What you share

Connecting grants the agent access to Streak on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Stripe
Source: https://getskydive.com/docs/integrations/stripe

Payment processing platform

## Example agent tasks

```text theme={null}
Connect Stripe and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Stripe.
```

```text theme={null}
Use the agent to interact with Stripe directly from your conversation.
```

## Connecting

Connect Stripe over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Stripe.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Stripe through the MCP server.

## What you share

Connecting grants the agent access to Stripe on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# SubscriptionFlow MCP server
Source: https://getskydive.com/docs/integrations/subscriptionflow

Subscription billing management

## Example agent tasks

```text theme={null}
Connect SubscriptionFlow MCP server and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from SubscriptionFlow MCP server.
```

```text theme={null}
Use the agent to interact with SubscriptionFlow MCP server directly from your conversation.
```

## Connecting

Connect SubscriptionFlow over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add SubscriptionFlow.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach SubscriptionFlow through the MCP server.

## What you share

Connecting grants the agent access to SubscriptionFlow MCP server on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Supabase
Source: https://getskydive.com/docs/integrations/supabase

Open source Postgres backend for apps

## Example agent tasks

```text theme={null}
Connect Supabase and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Supabase.
```

```text theme={null}
Use the agent to interact with Supabase directly from your conversation.
```

## Connecting

**MCP server**

Connect Supabase over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Supabase.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Supabase through the MCP server.

**API key**

Connect Supabase with an API key.

**1. Add the connection**

Generate an API key from your Supabase account settings.

**2. Authorize**

Copy the key. It is only shown once.

**3. The agent gets to work**

Paste it when prompted in the agent's connection card for Supabase.

## What you share

Connecting grants the agent access to Supabase on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Superhuman Mail
Source: https://getskydive.com/docs/integrations/superhuman

Fast email client for professionals

## Example agent tasks

```text theme={null}
Connect Superhuman Mail and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Superhuman Mail.
```

```text theme={null}
Use the agent to interact with Superhuman Mail directly from your conversation.
```

## Connecting

Connect Superhuman Mail over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Superhuman Mail.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Superhuman Mail through the MCP server.

## What you share

Connecting grants the agent access to Superhuman Mail on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Supermetrics Marketing Analytics
Source: https://getskydive.com/docs/integrations/supermetrics

Marketing data pipeline and reporting

## Example agent tasks

```text theme={null}
Connect Supermetrics Marketing Analytics and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Supermetrics Marketing Analytics.
```

```text theme={null}
Use the agent to interact with Supermetrics Marketing Analytics directly from your conversation.
```

## Connecting

Connect Supermetrics Marketing Analytics over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Supermetrics Marketing Analytics.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Supermetrics Marketing Analytics through the MCP server.

## What you share

Connecting grants the agent access to Supermetrics Marketing Analytics on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Sybill
Source: https://getskydive.com/docs/integrations/sybill

AI meeting notes for sales calls

## Example agent tasks

```text theme={null}
Connect Sybill and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Sybill.
```

```text theme={null}
Use the agent to interact with Sybill directly from your conversation.
```

## Connecting

Connect Sybill over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Sybill.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Sybill through the MCP server.

## What you share

Connecting grants the agent access to Sybill on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Synapse
Source: https://getskydive.com/docs/integrations/synapse

Biomedical research data collaboration platform

## Example agent tasks

```text theme={null}
Connect Synapse and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Synapse.
```

```text theme={null}
Use the agent to interact with Synapse directly from your conversation.
```

## Connecting

Connect Synapse over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Synapse.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Synapse through the MCP server.

## What you share

Connecting grants the agent access to Synapse on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Tally
Source: https://getskydive.com/docs/integrations/tally

Free online form and survey builder

## Example agent tasks

```text theme={null}
Connect Tally and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Tally.
```

```text theme={null}
Use the agent to interact with Tally directly from your conversation.
```

## Connecting

Connect Tally over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Tally.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Tally through the MCP server.

## What you share

Connecting grants the agent access to Tally on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Tana
Source: https://getskydive.com/docs/integrations/tana

Let your agent read and write your Tana knowledge graph

Tana is a structured note-taking and knowledge graph tool built around Supertags. Connect it and your agent can capture new nodes, search your graph, and pull structured data from your workspace.

## Two ways to connect

Tana offers both an MCP connection and a write-only Input API. For most use cases, MCP is the better choice since it supports both reading and writing.

* **Tana MCP (recommended).** Broad read and write access to your Tana graph. See [Connect with MCP](#connect-with-mcp).
* **Tana Input API.** Write-only access via a workspace token. Useful for pushing data into Tana when you don't need to read back. See [Connect with Input API](#connect-with-input-api).

## What the agent can do

> **Capture nodes**
>
> Create new nodes in your Tana graph with fields and Supertags applied.

> **Search your graph**
>
> Query nodes by tag, field value, or full-text search.

> **Read structured data**
>
> Pull node properties and relationships from your workspace.

> **Update nodes**
>
> Edit existing nodes, add fields, and apply or change tags.

## Example agent tasks

```text theme={null}
Add a new meeting note from today's call with Acme, tagged as a Meeting with the key action items filled in.
```

```text theme={null}
Find all nodes tagged as Project that are marked incomplete and list them.
```

```text theme={null}
Capture this research summary as a new node in my Reading List with the source and date.
```

## What you share

Connecting through MCP gives the agent access to your Tana graph within your workspace permissions. The Input API is scoped to write-only within the workspace you select. Both require a paid Tana plan. See [Permissions](/docs/integrations/permissions).

## Connect with MCP

**1. Add the Tana connection**

From the agent's connections, choose Tana. The platform connects you through Tana's remote MCP server.

**2. Sign in to Tana**

A connect card appears. Sign in to your Tana account and approve access. A paid Tana plan is required.

**3. The agent gets to work**

It can now read and write your Tana graph through the MCP server.

## Connect with Input API

**1. Get an Input API token**

In Tana Outliner, open your profile settings, go to API tokens, then Workspace tokens. Select a workspace and create a token.

**2. Add the Tana Input API connection**

From the agent's connections, choose Tana Input API and paste your token.

**3. The agent can write to your graph**

It can now push new nodes into your selected Tana workspace.

> [!NOTE]
>
> The Input API is write-only. Use MCP when you need the agent to read from your Tana graph as well.

## Learn more

> **[Tana documentation](https://tana.inc/docs)**
>
> Tana guides, the Supertag model, and the Input API reference.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Tavily
Source: https://getskydive.com/docs/integrations/tavily

Search API built for AI agents

## Example agent tasks

```text theme={null}
Connect Tavily and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Tavily.
```

```text theme={null}
Use the agent to interact with Tavily directly from your conversation.
```

## Connecting

Connect Tavily over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Tavily.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Tavily through the MCP server.

## What you share

Connecting grants the agent access to Tavily on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Teamwork
Source: https://getskydive.com/docs/integrations/teamwork

Let your agent manage projects and tasks in Teamwork.com

Teamwork.com is project management software for client work. Connect it and your agent can work with your projects, task lists, and tasks.

## What the agent can do

> **Read projects**
>
> See the projects you can access.

> **Manage tasks**
>
> Create, update, and complete tasks.

> **Work with task lists**
>
> Find and organize lists inside a project.

> **Track assignments**
>
> See who's on what and what's due.

## Example agent tasks

```text theme={null}
Each morning, list my open Teamwork tasks due this week grouped by project.
```

```text theme={null}
When a deal closes, create a project from our onboarding template and assign the kickoff tasks.
```

```text theme={null}
Mark these tasks done and move anything still blocked to next Monday.
```

## What you share

During the connect flow you authorize what the agent can reach. It works only within what you grant. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Authorize Teamwork for the agent**

From the agent's connections, choose Teamwork and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[Teamwork.com API docs](https://apidocs.teamwork.com/docs/teamwork)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# TickTick
Source: https://getskydive.com/docs/integrations/ticktick

Let your agent manage your TickTick tasks

TickTick is a to-do and habit app. Connect it and your agent can read your tasks and create or complete them.

## What the agent can do

> **Read tasks**
>
> See your tasks and lists.

> **Create tasks**
>
> Add tasks with due dates.

> **Complete tasks**
>
> Mark tasks done.

> **See lists**
>
> Find the right list.

## Example agent tasks

```text theme={null}
Turn my brain-dump into scheduled tasks across the week so nothing's forgotten.
```

```text theme={null}
Each morning, tell me the three things that actually matter today.
```

```text theme={null}
When I say I'm done, mark it and tee up what's next.
```

## What you share

During the connect flow you authorize what the agent can reach. It works only within what you grant. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Authorize TickTick for the agent**

From the agent's connections, choose TickTick and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[TickTick API docs](https://developer.ticktick.com/)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# pg-aiguide MCP server
Source: https://getskydive.com/docs/integrations/tigerdata

Postgres platform for AI and analytics

## Example agent tasks

```text theme={null}
Connect pg-aiguide MCP server and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from pg-aiguide MCP server.
```

```text theme={null}
Use the agent to interact with pg-aiguide MCP server directly from your conversation.
```

## Connecting

Connect pg-aiguide over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add pg-aiguide.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach pg-aiguide through the MCP server.

## What you share

Connecting grants the agent access to pg-aiguide MCP server on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Tigris
Source: https://getskydive.com/docs/integrations/tigrisdata

Globally distributed object storage

## Example agent tasks

```text theme={null}
Connect Tigris and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Tigris.
```

```text theme={null}
Use the agent to interact with Tigris directly from your conversation.
```

## Connecting

Connect Tigris over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Tigris.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Tigris through the MCP server.

## What you share

Connecting grants the agent access to Tigris on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Timely
Source: https://getskydive.com/docs/integrations/timely

Let your agent read your Timely time tracking

Timely automatically tracks time across your work. Connect it and your agent can read your tracked hours and projects.

## What the agent can do

> **Read time**
>
> See hours logged by project.

> **By project**
>
> Break down time per project.

> **By date**
>
> Read time over a range.

> **Summaries**
>
> Roll up your tracked hours.

## Example agent tasks

```text theme={null}
Each Friday, tell me where my week actually went and what to bill.
```

```text theme={null}
Flag when I've spent too long on low-value work this week.
```

```text theme={null}
Summarize my time on this client for the invoice.
```

## What you share

During the connect flow you authorize what the agent can reach. It works only within what you grant. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Authorize Timely for the agent**

From the agent's connections, choose Timely and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[Timely API docs](https://dev.timelyapp.com/)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Todoist
Source: https://getskydive.com/docs/integrations/todoist

Let your agent manage your Todoist tasks

Todoist is a personal and team to-do app. Connect it and your agent can read your tasks and create or complete them.

## What the agent can do

> **Read tasks**
>
> See your tasks by project or due date.

> **Create tasks**
>
> Add tasks with due dates.

> **Complete tasks**
>
> Mark tasks done.

> **See projects**
>
> Find the right project.

## Example agent tasks

```text theme={null}
Each morning, plan my day from what's due and what I keep pushing.
```

```text theme={null}
When I forward an email that needs action, turn it into a task with a deadline.
```

```text theme={null}
At the end of the week, tell me what I finished and what slipped.
```

## What you share

During the connect flow you authorize what the agent can reach. It works only within what you grant. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Authorize Todoist for the agent**

From the agent's connections, choose Todoist and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[Todoist API docs](https://developer.todoist.com/rest/v2/)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Transloadit hosted MCP server
Source: https://getskydive.com/docs/integrations/transloadit

File uploading and media processing infrastructure

## Example agent tasks

```text theme={null}
Connect Transloadit hosted MCP server and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Transloadit hosted MCP server.
```

```text theme={null}
Use the agent to interact with Transloadit hosted MCP server directly from your conversation.
```

## Connecting

Connect Transloadit hosted over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Transloadit hosted.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Transloadit hosted through the MCP server.

## What you share

Connecting grants the agent access to Transloadit hosted MCP server on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Trello
Source: https://getskydive.com/docs/integrations/trello

Let your agent manage your Trello boards and cards

Trello organizes work into boards, lists, and cards. Connect it and your agent can read your boards and create or move cards.

## What the agent can do

> **Read boards**
>
> See the lists and cards on a board.

> **Create cards**
>
> Add cards to a list.

> **Move cards**
>
> Move a card between lists.

> **Update cards**
>
> Edit card details and due dates.

## Example agent tasks

```text theme={null}
Turn each new inbound email into a card in my "Backlog" with the key details.
```

```text theme={null}
Every morning, move anything stale in "In Progress" to my attention.
```

```text theme={null}
When a card hits "Done," log it to our shipped-this-week list.
```

## What you share

During the connect flow you authorize what the agent can reach. It works only within what you grant. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Authorize Trello for the agent**

From the agent's connections, choose Trello and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[Trello REST docs](https://developer.atlassian.com/cloud/trello/rest/)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# KYC Document API / Customer API v3.0
Source: https://getskydive.com/docs/integrations/trulioo

Identity verification and KYC document checks

## Example agent tasks

```text theme={null}
Connect KYC Document API / Customer API v3.0 and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from KYC Document API / Customer API v3.0.
```

```text theme={null}
Use the agent to interact with KYC Document API / Customer API v3.0 directly from your conversation.
```

## Connecting

Connect KYC Document API / Customer API v3.0 with an API key.

**1. Add the connection**

Generate an API key from your KYC Document API / Customer API v3.0 account settings.

**2. Authorize**

Copy the key. It is only shown once.

**3. The agent gets to work**

Paste it when prompted in the agent's connection card for KYC Document API / Customer API v3.0.

## What you share

Connecting grants the agent access to KYC Document API / Customer API v3.0 on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Trustoo MCP Server
Source: https://getskydive.com/docs/integrations/trustoo

Customer review collection platform

## Example agent tasks

```text theme={null}
Connect Trustoo MCP Server and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Trustoo MCP Server.
```

```text theme={null}
Use the agent to interact with Trustoo MCP Server directly from your conversation.
```

## Connecting

Connect Trustoo over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Trustoo.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Trustoo through the MCP server.

## What you share

Connecting grants the agent access to Trustoo MCP Server on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Kopi
Source: https://getskydive.com/docs/integrations/trykopi

AI email design for Shopify marketing

## Example agent tasks

```text theme={null}
Connect Kopi and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Kopi.
```

```text theme={null}
Use the agent to interact with Kopi directly from your conversation.
```

## Connecting

Connect Kopi over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Kopi.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Kopi through the MCP server.

## What you share

Connecting grants the agent access to Kopi on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Tuist
Source: https://getskydive.com/docs/integrations/tuist

Xcode project generation and tooling

## Example agent tasks

```text theme={null}
Connect Tuist and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Tuist.
```

```text theme={null}
Use the agent to interact with Tuist directly from your conversation.
```

## Connecting

Connect Tuist over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Tuist.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Tuist through the MCP server.

## What you share

Connecting grants the agent access to Tuist on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Tumblr
Source: https://getskydive.com/docs/integrations/tumblr

Let your agent work with your Tumblr blog

Tumblr is a microblogging and social platform. Connect it and your agent can read your blog and help create posts.

## What the agent can do

> **Read posts**
>
> See your blog's posts.

> **Create posts**
>
> Publish new posts.

> **See engagement**
>
> Read likes and notes.

> **Find content**
>
> Look up your posts.

## Example agent tasks

```text theme={null}
Turn this update into a Tumblr post in my blog's voice and queue it.
```

```text theme={null}
Every week, tell me which posts got the most notes and why.
```

```text theme={null}
Find my best-performing post on this topic and suggest a follow-up.
```

## What you share

During the connect flow you authorize what the agent can reach. It works only within what you grant. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Authorize Tumblr for the agent**

From the agent's connections, choose Tumblr and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[Tumblr API docs](https://www.tumblr.com/docs/en/api/v2)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Twitch
Source: https://getskydive.com/docs/integrations/twitch

Let your agent work with your Twitch channel data

Twitch is live streaming for creators. Connect it and your agent can read your channel, streams, and followers.

## What the agent can do

> **Read streams**
>
> See your stream details and status.

> **Read followers**
>
> Pull follower and subscriber info.

> **See stats**
>
> Read viewer and channel stats.

> **Find content**
>
> Look up your videos and clips.

## Example agent tasks

```text theme={null}
After each stream, tell me how it went: viewers, new followers, and best moments.
```

```text theme={null}
When I go live, post the announcement to my other channels automatically.
```

```text theme={null}
Every week, tell me which streams grew my following the most.
```

## What you share

During the connect flow you authorize what the agent can reach. It works only within what you grant. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Authorize Twitch for the agent**

From the agent's connections, choose Twitch and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[Twitch API docs](https://dev.twitch.tv/docs/api/)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Typefully
Source: https://getskydive.com/docs/integrations/typefully

Social post writing and scheduling for X, LinkedIn, Threads, Bluesky, and Mastodon

Typefully is a social post writing and scheduling tool. Connect it and your agents can draft and refine posts in your voice, read and reply to draft comments, schedule or publish to X, LinkedIn, Threads, Bluesky, and Mastodon, and manage your content queue.

## What your agents can do

> **Draft and refine posts**
>
> Write drafts in your voice, read team feedback, and reply to draft comments as the post improves.

> **Schedule and publish**
>
> Schedule posts for specific times or publish immediately to X, LinkedIn, Threads, Bluesky, and Mastodon.

> **Manage your content queue**
>
> List scheduled posts, check draft status, and organize queued content with tags.

> **Handle media and accounts**
>
> Upload media to drafts and check which accounts are connected.

## Example agent tasks

```text theme={null}
Draft 5 posts from this article in my voice and schedule one per morning this week.
```

```text theme={null}
Check the comments on my latest draft, address the feedback, and update the post.
```

```text theme={null}
List everything queued for next week and flag posts missing a publish time.
```

## What you share

Connecting gives your agents access to your Typefully account through its MCP server: they can create and edit drafts, read and reply to draft comments, schedule and publish posts, manage queued content, upload media, and check connected accounts. See [Permissioning](/docs/integrations/permissions).

## Connecting

**MCP server**

Connect Typefully over MCP for direct access to its drafting and scheduling tools.

**1. Add the Typefully connection**

From your agent's connections, choose Typefully.

**2. Sign in to Typefully**

A connect card appears. Approve access with OAuth; Typefully asks for full access to your account.

**3. Your agents get to work**

They can now draft, schedule, and publish through Typefully's hosted MCP server at mcp.typefully.com.

## Learn more

> **[Typefully for AI agents](https://typefully.com/ai-agents)**
>
> How Typefully connects to AI tools, and what agents can do once connected.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Ubersuggest
Source: https://getskydive.com/docs/integrations/ubersuggest

Let your agent research keywords and SEO opportunities in Ubersuggest

Ubersuggest is an SEO and keyword research tool from Neil Patel. Connect it and your agent can pull keyword metrics, surface content ideas, and analyze your site's search performance.

## What the agent can do

> **Research keywords**
>
> Pull search volume, difficulty, and CPC data for any keyword.

> **Find content ideas**
>
> Discover related keywords and topics your audience is searching for.

> **Analyze competitors**
>
> See what keywords competitor sites rank for and what's driving their traffic.

> **Audit SEO health**
>
> Surface site issues, backlink data, and on-page optimization opportunities.

## Example agent tasks

```text theme={null}
What keywords should I target for a blog post about project management software? Include difficulty and volume.
```

```text theme={null}
Compare the keyword rankings for my site versus competitor.com and highlight gaps.
```

```text theme={null}
Find long-tail keyword opportunities in the email marketing space with low difficulty.
```

## What you share

Connecting gives the agent access to your Ubersuggest account and its keyword and site data. See [Permissions](/docs/integrations/permissions).

## Connecting

**1. Add the Ubersuggest connection**

From the agent's connections, choose Ubersuggest. The platform connects you through Ubersuggest's MCP server.

**2. Sign in to Ubersuggest**

A connect card appears. Sign in to your Ubersuggest account and approve access.

**3. The agent gets to work**

It can now pull keyword data, content ideas, and SEO analysis from Ubersuggest.

## Learn more

> **[Neil Patel Digital](https://app.neilpatel.com)**
>
> Ubersuggest app and account management.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Udemy Business
Source: https://getskydive.com/docs/integrations/udemy

Online courses for workplace learning

## Example agent tasks

```text theme={null}
Connect Udemy Business and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Udemy Business.
```

```text theme={null}
Use the agent to interact with Udemy Business directly from your conversation.
```

## Connecting

Connect Udemy (ChatGPT) over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Udemy Business.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Udemy Business through the MCP server.

## What you share

Connecting grants the agent access to Udemy Business on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# UptimeRobot
Source: https://getskydive.com/docs/integrations/uptimerobot

Website uptime monitoring

## Example agent tasks

```text theme={null}
Connect UptimeRobot and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from UptimeRobot.
```

```text theme={null}
Use the agent to interact with UptimeRobot directly from your conversation.
```

## Connecting

Connect UptimeRobot over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add UptimeRobot.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach UptimeRobot through the MCP server.

## What you share

Connecting grants the agent access to UptimeRobot on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Benefits Intake
Source: https://getskydive.com/docs/integrations/va

US veterans benefits claims document submission

## Example agent tasks

```text theme={null}
Connect Benefits Intake and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Benefits Intake.
```

```text theme={null}
Use the agent to interact with Benefits Intake directly from your conversation.
```

## Connecting

Connect Benefits Intake with an API key.

**1. Add the connection**

Generate an API key from your Benefits Intake account settings.

**2. Authorize**

Copy the key. It is only shown once.

**3. The agent gets to work**

Paste it when prompted in the agent's connection card for Benefits Intake.

## What you share

Connecting grants the agent access to Benefits Intake on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Vantage
Source: https://getskydive.com/docs/integrations/vantage

Cloud cost management and optimization

## Example agent tasks

```text theme={null}
Connect Vantage and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Vantage.
```

```text theme={null}
Use the agent to interact with Vantage directly from your conversation.
```

## Connecting

Connect Vantage over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Vantage.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Vantage through the MCP server.

## What you share

Connecting grants the agent access to Vantage on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Vectara REST API v2
Source: https://getskydive.com/docs/integrations/vectara

Retrieval-augmented and semantic search

## Example agent tasks

```text theme={null}
Connect Vectara REST API v2 and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Vectara REST API v2.
```

```text theme={null}
Use the agent to interact with Vectara REST API v2 directly from your conversation.
```

## Connecting

Connect Vectara REST API v2 with an API key.

**1. Add the connection**

Generate an API key from your Vectara REST API v2 account settings.

**2. Authorize**

Copy the key. It is only shown once.

**3. The agent gets to work**

Paste it when prompted in the agent's connection card for Vectara REST API v2.

## What you share

Connecting grants the agent access to Vectara REST API v2 on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Hindsight Cloud MCP server (multi-bank)
Source: https://getskydive.com/docs/integrations/vectorize

Persistent memory layer for AI agents

## Example agent tasks

```text theme={null}
Connect Hindsight Cloud MCP server (multi-bank) and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Hindsight Cloud MCP server (multi-bank).
```

```text theme={null}
Use the agent to interact with Hindsight Cloud MCP server (multi-bank) directly from your conversation.
```

## Connecting

Connect Hindsight Cloud (multi-bank) over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Hindsight Cloud MCP server (multi-bank).

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Hindsight Cloud MCP server (multi-bank) through the MCP server.

## What you share

Connecting grants the agent access to Hindsight Cloud MCP server (multi-bank) on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# VEED Fabric
Source: https://getskydive.com/docs/integrations/veed

Online video editing platform

## Example agent tasks

```text theme={null}
Connect VEED Fabric and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from VEED Fabric.
```

```text theme={null}
Use the agent to interact with VEED Fabric directly from your conversation.
```

## Connecting

Connect VEED Fabric over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add VEED Fabric.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach VEED Fabric through the MCP server.

## What you share

Connecting grants the agent access to VEED Fabric on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Venturu MCP server
Source: https://getskydive.com/docs/integrations/venturu

Startup and investor deal-flow data

## Example agent tasks

```text theme={null}
Connect Venturu MCP server and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Venturu MCP server.
```

```text theme={null}
Use the agent to interact with Venturu MCP server directly from your conversation.
```

## Connecting

Connect Venturu over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Venturu.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Venturu through the MCP server.

## What you share

Connecting grants the agent access to Venturu MCP server on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Vercel
Source: https://getskydive.com/docs/integrations/vercel

Let your agent work with your Vercel projects and deployments

Vercel hosts your frontends and serverless workloads. Connect it and your agent can read projects, check deployments, and work with your domains and teams.

## What the agent can do

> **List projects**
>
> See the projects in your Vercel team.

> **Check deployments**
>
> Read deployment status and recent builds.

> **Review domains**
>
> See domains attached to your projects.

> **Work with teams**
>
> Act in the team context you authorized.

## Example agent tasks

```text theme={null}
What's the status of the latest production deployment for my main project?
```

```text theme={null}
Every morning, summarize failed Vercel deployments from the last day and open a note with the error summary.
```

```text theme={null}
List my Vercel projects and which ones have a custom domain.
```

## What you share

During the connect flow you authorize what the agent can reach. It works only within what you grant. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Authorize Vercel for the agent**

From the agent's connections, choose Vercel and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[Vercel REST API docs](https://vercel.com/docs/rest-api)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Vimeo
Source: https://getskydive.com/docs/integrations/vimeo

Let your agent work with your Vimeo videos

Vimeo hosts your videos. Connect it and your agent can read your videos and their details.

## What the agent can do

> **Read videos**
>
> See your uploaded videos.

> **See stats**
>
> Pull view and engagement stats.

> **Browse**
>
> Move through your folders and showcases.

> **Find a video**
>
> Locate the right video.

## Example agent tasks

```text theme={null}
When a video crosses a view milestone, tell me and suggest where to promote it next.
```

```text theme={null}
Every Monday, tell me which videos are gaining traction.
```

```text theme={null}
Summarize the performance of this campaign's videos.
```

## What you share

During the connect flow you authorize what the agent can reach. It works only within what you grant. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Authorize Vimeo for the agent**

From the agent's connections, choose Vimeo and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[Vimeo API docs](https://developer.vimeo.com/api/reference)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Visualping
Source: https://getskydive.com/docs/integrations/visualping

Website change detection and monitoring

## Example agent tasks

```text theme={null}
Connect Visualping and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Visualping.
```

```text theme={null}
Use the agent to interact with Visualping directly from your conversation.
```

## Connecting

Connect Visualping over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Visualping.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Visualping through the MCP server.

## What you share

Connecting grants the agent access to Visualping on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# WakaTime
Source: https://getskydive.com/docs/integrations/wakatime

Let your agent read your WakaTime coding stats

WakaTime tracks time spent coding. Connect it and your agent can read your coding activity and summaries.

## What the agent can do

> **Read time stats**
>
> See time spent by project or language.

> **Summaries**
>
> Pull daily and weekly summaries.

> **By project**
>
> Break down activity per project.

> **By date**
>
> Read stats for a date range.

## Example agent tasks

```text theme={null}
Every Friday, tell me where my coding time actually went this week.
```

```text theme={null}
Flag when I've spent a whole day in meetings with no focus time.
```

```text theme={null}
Summarize which project ate the most hours this month.
```

## What you share

During the connect flow you authorize what the agent can reach. It works only within what you grant. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Authorize WakaTime for the agent**

From the agent's connections, choose WakaTime and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[WakaTime API docs](https://wakatime.com/developers)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# W&B
Source: https://getskydive.com/docs/integrations/wandb

Experiment tracking for machine learning

## Example agent tasks

```text theme={null}
Connect W&B and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from W&B.
```

```text theme={null}
Use the agent to interact with W&B directly from your conversation.
```

## Connecting

Connect W\&B over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add W\&B.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach W\&B through the MCP server.

## What you share

Connecting grants the agent access to W\&B on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Webex
Source: https://getskydive.com/docs/integrations/webex

Let your agent read and post in Webex

Webex runs meetings and team messaging. Connect it and your agent can read messages and post to your spaces.

## What the agent can do

> **Read messages**
>
> Catch up on a Webex space.

> **Post messages**
>
> Send updates or replies.

> **See meetings**
>
> Read your scheduled meetings.

> **See spaces**
>
> Find the right space.

## Example agent tasks

```text theme={null}
Each morning, summarize the overnight activity in my team spaces.
```

```text theme={null}
After a meeting, post the action items to the space and assign owners.
```

```text theme={null}
Remind the space about the next meeting an hour before it starts.
```

## What you share

During the connect flow you authorize what the agent can reach. It works only within what you grant. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Authorize Webex for the agent**

From the agent's connections, choose Webex and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[Webex API docs](https://developer.webex.com/docs)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Webflow
Source: https://getskydive.com/docs/integrations/webflow

Let your agent work with your Webflow site content

Webflow builds sites with a visual editor and CMS. Connect it and your agent can read and update your CMS content.

## What the agent can do

> **Read CMS items**
>
> Pull entries from your collections.

> **Create items**
>
> Add new CMS entries.

> **Update items**
>
> Edit existing entries.

> **See sites**
>
> Find the right site and collection.

## Example agent tasks

```text theme={null}
When a case study is approved, publish it to the CMS with the right fields.
```

```text theme={null}
Every Monday, tell me which CMS entries are stale and need a refresh.
```

```text theme={null}
Add this new customer logo and quote to the testimonials collection.
```

## What you share

During the connect flow you authorize what the agent can reach. It works only within what you grant. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Add the Webflow connection**

From the agent's connections, choose Webflow. The platform connects you through Webflow's MCP server.

**2. Sign in to Webflow**

A connect card appears. Sign in to your Webflow account and approve access to your sites and CMS.

**3. The agent gets to work**

It can now read and update your CMS content, manage collections, and work across your Webflow sites.

## Learn more

> **[Webflow API docs](https://developers.webflow.com/data/reference)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Webhook Relay
Source: https://getskydive.com/docs/integrations/webhookrelay

Webhook forwarding and relay service

## Example agent tasks

```text theme={null}
Connect Webhook Relay and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Webhook Relay.
```

```text theme={null}
Use the agent to interact with Webhook Relay directly from your conversation.
```

## Connecting

Connect Webhook Relay with an API key.

**1. Add the connection**

Generate an API key from your Webhook Relay account settings.

**2. Authorize**

Copy the key. It is only shown once.

**3. The agent gets to work**

Paste it when prompted in the agent's connection card for Webhook Relay.

## What you share

Connecting grants the agent access to Webhook Relay on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Wednesday.app MCP
Source: https://getskydive.com/docs/integrations/wednesday

Influencer marketing and creator discovery

## Example agent tasks

```text theme={null}
Connect Wednesday.app MCP and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Wednesday.app MCP.
```

```text theme={null}
Use the agent to interact with Wednesday.app MCP directly from your conversation.
```

## Connecting

Connect Wednesday.app over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Wednesday.app.app MCP's MCP server.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Wednesday.app MCP through the MCP server.

## What you share

Connecting grants the agent access to Wednesday.app MCP on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Whimsical
Source: https://getskydive.com/docs/integrations/whimsical

Visual collaboration for flowcharts and diagrams

## Example agent tasks

```text theme={null}
Connect Whimsical and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Whimsical.
```

```text theme={null}
Use the agent to interact with Whimsical directly from your conversation.
```

## Connecting

Connect Whimsical over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Whimsical.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Whimsical through the MCP server.

## What you share

Connecting grants the agent access to Whimsical on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Windmill
Source: https://getskydive.com/docs/integrations/windmill

Open-source workflow automation platform

## Example agent tasks

```text theme={null}
Connect Windmill and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Windmill.
```

```text theme={null}
Use the agent to interact with Windmill directly from your conversation.
```

## Connecting

Connect Windmill with an API key.

**1. Add the connection**

Generate an API key from your Windmill account settings.

**2. Authorize**

Copy the key. It is only shown once.

**3. The agent gets to work**

Paste it when prompted in the agent's connection card for Windmill.

## What you share

Connecting grants the agent access to Windmill on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Windsor
Source: https://getskydive.com/docs/integrations/windsor

Let your agent pull and unify marketing data from across your ad channels in Windsor

Windsor is a no-code marketing data integration and reporting platform. Connect it and your agent can pull performance data from ad channels, build unified reports, and surface attribution insights across your campaigns.

## What the agent can do

> **Pull channel data**
>
> Retrieve performance metrics from Google Ads, Facebook Ads, and other connected ad platforms.

> **Build unified reports**
>
> Combine data across channels into a single view for campaign and budget analysis.

> **Surface attribution insights**
>
> Trace conversions back to their sources and identify which campaigns are driving results.

> **Query your data**
>
> Run custom queries against your Windsor data warehouse to answer specific marketing questions.

## Example agent tasks

```text theme={null}
Show last week's ROAS for each of our active Google Ads campaigns.
```

```text theme={null}
Compare Facebook and Google spend versus conversions for the past 30 days.
```

```text theme={null}
Which campaign drove the most revenue last month, and what was the CPA?
```

## What you share

During the connect flow you sign in to your Windsor account. The agent works within what you grant. The token is stored securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Add the Windsor connection**

From the agent's connections, choose Windsor. The platform connects through Windsor's MCP server.

**2. Sign in to Windsor**

A connect card appears. Sign in to your Windsor account and approve access.

**3. The agent gets to work**

It can now pull channel data, build reports, and surface attribution insights from your marketing data.

## Learn more

> **[Windsor docs](https://windsor.ai/docs)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Wispr Flow
Source: https://getskydive.com/docs/integrations/wispr-flow

AI notetaker for meetings - read your recordings, transcripts, and notes

Wispr Flow records and transcribes your meetings, producing searchable transcripts, summaries, and notes. Connect it and your agent can pull meeting content, review past recordings, and surface calendar context without you having to dig through recordings yourself.

## What the agent can do

> **Read meeting transcripts**
>
> Retrieve the full transcript from any recorded meeting for review or follow-up.

> **Pull meeting summaries**
>
> Get AI-generated summaries of meetings so the agent can brief you on what was covered.

> **Access notes**
>
> Read notes attached to meetings and recordings in your Wispr Flow workspace.

> **Review calendar context**
>
> Surface attendee details and calendar events linked to your meetings.

## Example agent tasks

```text theme={null}
Summarize the key decisions from my last three meetings this week.
```

```text theme={null}
Pull the transcript from yesterday's product review call and list any open questions.
```

```text theme={null}
Who attended the design sync on Monday and what was discussed?
```

## What you share

Connecting gives the agent read-only access to your Wispr Flow account: meeting transcripts, summaries, notes, attendee emails, and calendar events. It cannot use Wispr's dictation or voice-model APIs and cannot create, modify, or delete anything. Requires a Wispr plan with Notetaker. Not available on Enterprise or HIPAA accounts. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Add the Wispr Flow connection**

From the agent's connections, choose Wispr Flow. The platform connects through Wispr's MCP server.

**2. Sign in to Wispr Flow**

A connect card appears. Sign in to your Wispr Flow account and approve access.

**3. The agent gets to work**

It can now read your meeting transcripts, summaries, and notes from your Wispr Flow workspace.

## Learn more

> **[Wispr Flow support](https://support.wisprflow.ai)**
>
> Product guides and help documentation.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Wix
Source: https://getskydive.com/docs/integrations/wix

Website builder with a large app market and add-ons

## Example agent tasks

```text theme={null}
Connect Wix and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Wix.
```

```text theme={null}
Use the agent to interact with Wix directly from your conversation.
```

## Connecting

Connect Wix over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Wix.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Wix through the MCP server.

## What you share

Connecting grants the agent access to Wix on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# WordPress.com
Source: https://getskydive.com/docs/integrations/wordpress

Let your agent read and write your WordPress posts

WordPress.com hosts your sites and blog. Connect it and your agent can create posts, manage pages, update settings, and work with your media library.

## Two ways to connect

WordPress.com now runs a hosted MCP server. Connect it and the agent can work across your sites and content.

* **WordPress.com MCP (recommended).** Connect WordPress.com's MCP server once and the agent can create and edit posts and pages, manage media, update site settings, handle comments. See [Connect with MCP](#connect-with-mcp).
* **WordPress.com OAuth.** The original sign-in integration, scoped to what you grant. See [Connect with OAuth](#connect-with-oauth).

## What the agent can do

> **Create and edit posts**
>
> Write, update, and publish blog posts.

> **Manage pages**
>
> Create and edit static pages on your site.

> **Update site settings**
>
> Change your site title, tagline, and other settings.

> **Work with media**
>
> Upload and manage files in your media library.

## Example agent tasks

```text theme={null}
Write and save a draft post about our latest product update in my usual style.
```

```text theme={null}
List all posts that have been in draft for more than a month.
```

```text theme={null}
Create a new About page with a short bio from the notes I shared.
```

## Connect with MCP

**1. Add the WordPress.com connection**

From the agent's connections, choose WordPress.com. The platform connects you through WordPress.com's MCP server.

**2. Sign in to WordPress.com**

A connect card appears. Sign in to your WordPress.com account and approve.

**3. The agent gets to work**

It can now create and edit posts and pages, manage media, update site settings, handle comments through the MCP server.

> [!NOTE]
>
> MCP is the first-class connection when both options are available. See [Tools & MCP](/docs/capabilities/tools-and-mcp).

## Connect with OAuth

## What you share

During the connect flow you authorize what the agent can reach. It works only within what you grant. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

**1. Authorize WordPress.com for the agent**

From the agent's connections, choose WordPress.com and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[WordPress.com API docs](https://developer.wordpress.com/docs/api)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Workato
Source: https://getskydive.com/docs/integrations/workato

Let your agent run enterprise integrations and workflows in Workato

Workato is an enterprise integration and automation platform used across IT, finance, HR, and ops teams. Connect it and your agent can invoke recipes, move data between systems, and surface workflow results through a single conversation.

## What the agent can do

> **Invoke recipes**
>
> Start Workato recipes on demand from a conversation or triggered by another step.

> **Read job results**
>
> Pull the output and status of completed automation jobs.

> **Manage connections**
>
> Inspect and work with your Workato connection configuration.

> **Work with data**
>
> Pass and transform data across Workato-connected apps and systems.

## Example agent tasks

```text theme={null}
Run the new-hire provisioning recipe for the employee I just added in Workday.
```

```text theme={null}
Check whether last night's finance reconciliation job completed without errors.
```

```text theme={null}
Trigger the CRM sync and tell me how many records were updated.
```

## What you share

Connecting gives the agent access to your Workato workspace via the Developer API. Access is scoped to your API client's permissions. See [Permissions](/docs/integrations/permissions).

## Connecting

**1. Create a Workato API client**

In Workato, go to Workspace admin, then API clients. Create a client and copy the token. This is available on certain Workato plans.

**2. Add the Workato connection**

From the agent's connections, choose Workato and enter your API token when prompted.

**3. The agent gets to work**

It can now invoke recipes and read automation results from your Workato workspace.

## Learn more

> **[Workato documentation](https://docs.workato.com)**
>
> Full platform docs and the Developer API reference.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Workos
Source: https://getskydive.com/docs/integrations/workos

Enterprise SSO and auth infrastructure for developers

## Example agent tasks

```text theme={null}
Connect Workos and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Workos.
```

```text theme={null}
Use the agent to interact with Workos directly from your conversation.
```

## Connecting

Connect Workos over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Workos.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Workos through the MCP server.

## What you share

Connecting grants the agent access to Workos on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# X
Source: https://getskydive.com/docs/integrations/x

Social network for real-time posts

## Example agent tasks

```text theme={null}
Connect X and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from X.
```

```text theme={null}
Use the agent to interact with X directly from your conversation.
```

## Connecting

Connect X over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add X.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach X through the MCP server.

## What you share

Connecting grants the agent access to X on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Xero
Source: https://getskydive.com/docs/integrations/xero

Let your agent work with your Xero accounting

Xero is cloud accounting for small businesses. Connect it and your agent can read invoices, contacts, and reports, and create records.

## What the agent can do

> **Read invoices**
>
> See invoices and bills.

> **Read contacts**
>
> Pull customers and suppliers.

> **Create records**
>
> Add invoices and bills.

> **See reports**
>
> Read financial summaries.

## Example agent tasks

```text theme={null}
Each week, chase the unpaid invoices with a friendly nudge to each client.
```

```text theme={null}
When a bill comes in, categorize it and flag anything unusual.
```

```text theme={null}
Give me a plain-English read on this quarter's profit and what drove it.
```

## What you share

During the connect flow you authorize what the agent can reach. It works only within what you grant. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Authorize Xero for the agent**

From the agent's connections, choose Xero and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[Xero API docs](https://developer.xero.com/documentation/api/accounting/overview)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Yahoo
Source: https://getskydive.com/docs/integrations/yahoo

Let your agent work with your connected Yahoo account

Yahoo offers mail and other services. Connect it and your agent can work with the Yahoo data you authorize.

## What the agent can do

> **Read mail**
>
> Access the mail you grant.

> **Contacts**
>
> Look up your contacts.

> **Search**
>
> Find what you need.

> **Other data**
>
> Reach the Yahoo services you allow.

## Example agent tasks

```text theme={null}
Each morning, summarize my Yahoo inbox and surface what needs a reply.
```

```text theme={null}
Find the email with the details I need for this booking.
```

```text theme={null}
Flag anything time-sensitive sitting in my inbox.
```

## What you share

During the connect flow you authorize what the agent can reach. It works only within what you grant. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Authorize Yahoo for the agent**

From the agent's connections, choose Yahoo and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[Yahoo developer docs](https://developer.yahoo.com/)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# YNAB
Source: https://getskydive.com/docs/integrations/youneedabudget

Personal budgeting and money management

## Example agent tasks

```text theme={null}
Connect YNAB and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from YNAB.
```

```text theme={null}
Use the agent to interact with YNAB directly from your conversation.
```

## Connecting

Connect YNAB with an API key.

**1. Add the connection**

Generate an API key from your YNAB account settings.

**2. Authorize**

Copy the key. It is only shown once.

**3. The agent gets to work**

Paste it when prompted in the agent's connection card for YNAB.

## What you share

Connecting grants the agent access to YNAB on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Zapier
Source: https://getskydive.com/docs/integrations/zapier

No-code workflow automation between apps

## Example agent tasks

```text theme={null}
Connect Zapier and have the agent explore what it can help with.
```

```text theme={null}
Ask the agent to retrieve and summarize information from Zapier.
```

```text theme={null}
Use the agent to interact with Zapier directly from your conversation.
```

## Connecting

Connect Zapier over MCP for direct access to its tools.

**1. Add the connection**

From the agent's tools, add Zapier.

**2. Authorize**

A connect card appears. Authorize and approve access.

**3. The agent gets to work**

The agent can now reach Zapier through the MCP server.

## What you share

Connecting grants the agent access to Zapier on your behalf. As with every connection, credentials are handled by the secure [proxy](/docs/integrations/permissions) and never exposed to the model.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the model's reach.


# Zoho
Source: https://getskydive.com/docs/integrations/zoho

Let your agent work across your Zoho account

Zoho is a suite of business apps under one sign-in. Connect it and your agent can reach the Zoho apps you authorize.

## What the agent can do

> **Reach your apps**
>
> Work across the Zoho apps you grant.

> **CRM and contacts**
>
> Read records and contacts.

> **Finance**
>
> Reach invoicing and books data.

> **Find content**
>
> Search across your Zoho data.

## Example agent tasks

```text theme={null}
Connect the dots across my Zoho apps: when a deal closes in CRM, invoice it in Books and start onboarding.
```

```text theme={null}
Every Monday, give me one rollup across CRM, Books, and Desk of what needs attention.
```

```text theme={null}
Trace this customer from first contact to last invoice across my Zoho apps.
```

## What you share

During the connect flow you authorize what the agent can reach. It works only within what you grant. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Authorize Zoho for the agent**

From the agent's connections, choose Zoho and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[Zoho developer docs](https://www.zoho.com/developer/)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Zoho Bigin
Source: https://getskydive.com/docs/integrations/zoho-bigin

Let your agent work with your Zoho Bigin pipeline

Zoho Bigin is a simple CRM for small businesses. Connect it and your agent can read contacts and deals, and update them.

## What the agent can do

> **Read records**
>
> Pull contacts and deals.

> **Create records**
>
> Add contacts and deals.

> **Update records**
>
> Change fields and stages.

> **Search**
>
> Find the right record.

## Example agent tasks

```text theme={null}
When a lead arrives, add it and schedule the first follow-up.
```

```text theme={null}
Each Friday, tell me which deals stalled this week.
```

```text theme={null}
Move deals that closed into the onboarding flow and notify me.
```

## What you share

During the connect flow you authorize what the agent can reach. It works only within what you grant. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Authorize Zoho Bigin for the agent**

From the agent's connections, choose Zoho Bigin and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[Zoho Bigin API docs](https://www.bigin.com/developer/docs/apis/)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Zoho Books
Source: https://getskydive.com/docs/integrations/zoho-books

Let your agent work with your Zoho Books accounting

Zoho Books is online accounting. Connect it and your agent can read invoices, contacts, and expenses, and create records.

## What the agent can do

> **Read invoices**
>
> See invoices and their status.

> **Read contacts**
>
> Pull customers and vendors.

> **Create records**
>
> Add invoices and expenses.

> **See reports**
>
> Read financial summaries.

## Example agent tasks

```text theme={null}
Every Monday, chase overdue invoices with a friendly nudge.
```

```text theme={null}
When a bill arrives, categorize it and flag anything unusual.
```

```text theme={null}
Summarize this quarter's finances in plain English.
```

## What you share

During the connect flow you authorize what the agent can reach. It works only within what you grant. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Authorize Zoho Books for the agent**

From the agent's connections, choose Zoho Books and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[Zoho Books API docs](https://www.zoho.com/books/api/v3/)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Zoho Calendar
Source: https://getskydive.com/docs/integrations/zoho-calendar

Let your agent work with your Zoho Calendar

Zoho Calendar manages your schedule. Connect it and your agent can read your events and create new ones.

## What the agent can do

> **Read events**
>
> See what's on your calendar.

> **Create events**
>
> Add new events and meetings.

> **Check availability**
>
> See when you're free or busy.

> **Find events**
>
> Search your schedule.

## Example agent tasks

```text theme={null}
Each morning, brief me on my day and surface any conflicts.
```

```text theme={null}
Schedule this at the next slot we're both free.
```

```text theme={null}
Tell me where my week is overbooked.
```

## What you share

During the connect flow you authorize what the agent can reach. It works only within what you grant. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Authorize Zoho Calendar for the agent**

From the agent's connections, choose Zoho Calendar and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[Zoho Calendar API docs](https://www.zoho.com/calendar/help/api/)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Zoho CRM
Source: https://getskydive.com/docs/integrations/zoho-crm

Let your agent work with your Zoho CRM

Zoho CRM runs your sales pipeline. Connect it and your agent can read leads, contacts, and deals, and update records.

## What the agent can do

> **Read records**
>
> Pull leads, contacts, and deals.

> **Create records**
>
> Add new leads and contacts.

> **Update records**
>
> Change fields and move deals.

> **Search**
>
> Find the right record.

## Example agent tasks

```text theme={null}
When a deal closes, trigger onboarding and alert the account owner.
```

```text theme={null}
Every Monday, give me my pipeline with what's slipping and what to chase.
```

```text theme={null}
Enrich and route each new inbound lead to the right rep automatically.
```

## What you share

During the connect flow you authorize what the agent can reach. It works only within what you grant. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Authorize Zoho CRM for the agent**

From the agent's connections, choose Zoho CRM and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[Zoho CRM API docs](https://www.zoho.com/crm/developer/docs/api/v8/)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Zoho Desk
Source: https://getskydive.com/docs/integrations/zoho-desk

Let your agent work with your Zoho Desk support tickets

Zoho Desk runs customer support. Connect it and your agent can read tickets and help respond.

## What the agent can do

> **Read tickets**
>
> See open and recent tickets.

> **Read contacts**
>
> Pull a customer's details.

> **Draft replies**
>
> Prepare responses for agents.

> **Search**
>
> Find the right ticket.

## Example agent tasks

```text theme={null}
Every morning, triage new tickets and draft replies to the common ones.
```

```text theme={null}
When a ticket reopens, escalate it to me with the full history.
```

```text theme={null}
Every Friday, tell me the top issues customers raised this week.
```

## What you share

During the connect flow you authorize what the agent can reach. It works only within what you grant. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Authorize Zoho Desk for the agent**

From the agent's connections, choose Zoho Desk and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[Zoho Desk API docs](https://desk.zoho.com/DeskAPIDocument)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Zoho Inventory
Source: https://getskydive.com/docs/integrations/zoho-inventory

Let your agent work with your Zoho Inventory stock

Zoho Inventory manages stock and orders. Connect it and your agent can read items, stock levels, and orders.

## What the agent can do

> **Read items**
>
> See your products and stock levels.

> **Read orders**
>
> Pull sales and purchase orders.

> **Track shipments**
>
> See what's shipping.

> **Find items**
>
> Search your inventory.

## Example agent tasks

```text theme={null}
Each morning, flag anything low on stock before it runs out.
```

```text theme={null}
When an order ships, update stock and notify the customer.
```

```text theme={null}
Tell me which products are moving and which are sitting.
```

## What you share

During the connect flow you authorize what the agent can reach. It works only within what you grant. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Authorize Zoho Inventory for the agent**

From the agent's connections, choose Zoho Inventory and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[Zoho Inventory API docs](https://www.zoho.com/inventory/api/v1/)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Zoho Invoice
Source: https://getskydive.com/docs/integrations/zoho-invoice

Let your agent work with your Zoho Invoice billing

Zoho Invoice handles invoicing for small businesses. Connect it and your agent can read invoices and customers, and create invoices.

## What the agent can do

> **Read invoices**
>
> See invoices and their status.

> **Read customers**
>
> Pull customer details.

> **Create invoices**
>
> Add new invoices.

> **Track payments**
>
> See what's paid and outstanding.

## Example agent tasks

```text theme={null}
When a project wraps, generate and send the invoice automatically.
```

```text theme={null}
Each week, tell me what's unpaid and who to chase.
```

```text theme={null}
Summarize what I billed this month versus last.
```

## What you share

During the connect flow you authorize what the agent can reach. It works only within what you grant. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Authorize Zoho Invoice for the agent**

From the agent's connections, choose Zoho Invoice and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[Zoho Invoice API docs](https://www.zoho.com/invoice/api/v3/)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Zoho Mail
Source: https://getskydive.com/docs/integrations/zoho-mail

Let your agent work with your Zoho Mail

Zoho Mail is business email. Connect it and your agent can read your mail and help draft replies.

## What the agent can do

> **Read mail**
>
> Catch up on your inbox.

> **Draft replies**
>
> Prepare responses to send.

> **Search**
>
> Find a specific email.

> **Browse folders**
>
> Move through your mail folders.

## Example agent tasks

```text theme={null}
Each morning, summarize my inbox and draft replies to the routine ones.
```

```text theme={null}
Find the email with the contract and pull out the key dates.
```

```text theme={null}
Flag anything urgent that's been sitting unanswered.
```

## What you share

During the connect flow you authorize what the agent can reach. It works only within what you grant. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Authorize Zoho Mail for the agent**

From the agent's connections, choose Zoho Mail and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[Zoho Mail API docs](https://www.zoho.com/mail/help/api/)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Zoho People
Source: https://getskydive.com/docs/integrations/zoho-people

Let your agent work with your Zoho People HR data

Zoho People runs HR and team management. Connect it and your agent can read employee records and time-off data.

## What the agent can do

> **Read employees**
>
> See your team's records.

> **Time off**
>
> Read leave and attendance.

> **Search**
>
> Find an employee.

> **Summaries**
>
> Roll up HR data.

## Example agent tasks

```text theme={null}
Each Monday, tell me who's out this week so I can plan around it.
```

```text theme={null}
When someone joins, kick off their onboarding checklist.
```

```text theme={null}
Flag anyone whose leave balance is running low.
```

## What you share

During the connect flow you authorize what the agent can reach. It works only within what you grant. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Authorize Zoho People for the agent**

From the agent's connections, choose Zoho People and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[Zoho People API docs](https://www.zoho.com/people/api/)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Zoho Recruit
Source: https://getskydive.com/docs/integrations/zoho-recruit

Let your agent work with your Zoho Recruit hiring data

Zoho Recruit runs recruiting and applicant tracking. Connect it and your agent can read candidates and job openings.

## What the agent can do

> **Read candidates**
>
> Pull candidate profiles and status.

> **Read openings**
>
> See your job openings.

> **Update records**
>
> Move candidates through stages.

> **Search**
>
> Find the right candidate.

## Example agent tasks

```text theme={null}
When a strong candidate applies, flag them and draft the screening invite.
```

```text theme={null}
Every Monday, tell me which roles are stuck and where candidates are dropping off.
```

```text theme={null}
Summarize this candidate against the role before the interview.
```

## What you share

During the connect flow you authorize what the agent can reach. It works only within what you grant. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Authorize Zoho Recruit for the agent**

From the agent's connections, choose Zoho Recruit and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[Zoho Recruit API docs](https://www.zoho.com/recruit/developer-guide/apiv2/)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Zoom
Source: https://getskydive.com/docs/integrations/zoom

Let your agent work with your Zoom meetings and recordings

Zoom runs video meetings and webinars. Connect it and your agents can read your meetings, recordings, and transcripts.

## What your agents can do

> **Read meetings**
>
> See your scheduled and past meetings.

> **Recordings**
>
> Pull recordings and transcripts.

> **Schedule**
>
> Read and update upcoming meetings.

> **See participants**
>
> Read who attended.

## Example agent tasks

```text theme={null}
After each call, send me the summary, decisions, and action items.
```

```text theme={null}
Every Friday, pull the week's call notes into one recap.
```

```text theme={null}
When I miss a meeting, summarize the recording so I'm caught up in two minutes.
```

## What you share

During the connect flow you authorize what the agent can reach. It works only within what you grant. As with every connection, the underlying token is handled securely and never exposed to the model. See [Permissioning](/docs/integrations/permissions).

## Connecting

**1. Authorize Zoom for the agent**

From the agent's connections, choose Zoom and sign in.

**2. Approve the access**

Confirm what the agent can reach. You can change this later.

**3. The agent gets to work**

It works through the secure [proxy](/docs/integrations/permissions), within what you granted.

## Learn more

> **[Zoom API docs](https://developers.zoom.us/docs/api/)**
>
> The full reference for what's possible.

> **[How credentials stay safe](/docs/integrations/permissions)**
>
> The proxy model that keeps tokens out of the agent's reach.


# Support
Source: https://getskydive.com/docs/support

Get help with Skydive: email our team or find quick answers in the docs.

Need a hand with Skydive? Email [support@skydive.com](mailto:support@skydive.com) and we'll help. For quick answers, the pages below cover the most common questions.

## Email us

For anything the docs don't answer, including bugs and feature requests, email [support@skydive.com](mailto:support@skydive.com). To get a useful reply fast, tell us:

* What you were trying to do
* What happened instead, and what you expected
* The name of the agent involved, with a link to the conversation if you have one

## Find answers fast

Most questions are already covered in the docs.

> **[Quickstart](/docs/get-started/quickstart)**
>
> Set up your first agent

> **[FAQ](/docs/get-started/faq)**
>
> Common questions, answered

> **[Channels](/docs/channels/overview)**
>
> Reach your agent anywhere

> **[Pricing](/docs/account/pricing)**
>
> Plans and what's included


# Agents working together
Source: https://getskydive.com/docs/workspaces/agent-teams

How to build a team of agents that hands off, reviews, and escalates across your workspace

![A team of agents collaborating across a workspace](https://mintcdn.com/skydive/7-08Nfi5G-NOgAJr/images/agents/agent-teams.png?fit=max\&auto=format\&n=7-08Nfi5G-NOgAJr\&q=85\&s=e53c4064f4b4245c271fc51aef502a13)

Your workspace can hold as many agents as you need, each with its own job, connected to the right tools and channels, trained on what it needs to know. When you wire them together, they cover ground no single agent can.

This is different from [subagents](/docs/capabilities/subagents), which is one agent splitting its own work into parallel pieces. This is about multiple purpose-built agents handing off to each other across your team.

## Why more than one agent

One agent handles most tasks fine. You want a team when the work benefits from keeping things separate.

**Tooling access by role.** Each agent connects only to the tools its job requires. A CX agent stays connected to your support channels. A coding agent connects to GitHub and CI. They can work the same problem from both sides, visible to each other in a shared channel, while each acts only through its own toolset. The boundary is about what they act on, not what they know about, and the collaboration between them is where the real output comes from.

**Skill sets and training.** An agent trained entirely on your support history reasons about customer problems differently than one trained on your engineering codebase. Mixing both into one agent means neither specialization runs deep. Separate agents let each one develop genuine expertise. The more focused the agent's training, the better it gets at its job.

**The right model for the right job.** Not every task needs the same horsepower. A CX agent handling tier-1 questions runs well on a mid-tier model: fast, cost-effective, and sharp enough for the work. A coding agent doing deep reasoning over a complex PR belongs on a flagship model like Claude Fable 5 or Opus 4.8, where the quality actually matters. Splitting agents by role means you can match the model to the task instead of paying flagship rates for everything or underserving the jobs that need it.

**Review and escalation.** Some work needs a second opinion before it goes anywhere. An agent that handles inbound questions can pass work to an internal agent whose job is to check before anything reaches a human or touches a system.

**Parallel lanes.** Multiple agents working at once finish faster than one agent juggling everything. One reads the thread, one checks the system, one drafts the reply.

## Group channels and shared conversations

Agents can join group channels in Slack, just like a teammate would. This is where multi-agent work often happens naturally: a CX agent in your customer-facing channel, a coding agent in `#eng`, and humans pulling them both into threads when a problem spans both lanes. They collaborate in the open, each acting through its own tools, toward a shared outcome.

Group channels are also how you train agents to collaborate well. When an agent sits in a channel where handoffs, reviews, and escalations play out in real conversations, it learns the patterns firsthand. It picks up your team's voice, your escalation signals, when to act versus when to loop someone in. That context builds up over time and makes the agent more useful in its specific lane.

A few things to keep in mind for group channels:

* **Agents read the full channel history when they join.** They'll catch up on context from existing conversations. Be thoughtful about what's in a channel before adding an agent.
* **Each agent responds only when addressed or triggered.** Adding an agent to a channel doesn't make it interject on every message. It waits to be pulled in.
* **The agent knows who's talking.** Skydive recognizes each participant by their Slack identity, so the agent can tailor how it responds to different teammates while drawing on shared memory.

## Training for a specific lane

Siloing tooling between agents isn't just about access control. It's one of the best levers you have for getting an agent to develop real depth.

When one agent handles all of your customer questions, all of your engineering questions, and all of your scheduling, its training spreads across everything and goes deep on nothing. Split it: a CX agent builds up months of support patterns, escalation signals, product knowledge, and customer context. An engineering agent accumulates your codebase's conventions, incident history, and PR feedback. Each one gets better, faster, at its specific job, and when they work together in a shared channel, the output is better than either could produce alone.

The practical side: train each agent in the channel or conversation where its actual work happens. A CX agent trained in your support threads and customer Slack connect learns things a CX agent trained on general docs never will. An agent sitting in `#eng-incidents` for six months knows which systems fail together, which fixes work, and which engineers to loop in. That's not something you write into a system prompt. It accumulates.

## A real example: Anything's support team

The clearest illustration of this running in practice is Anything's own customer support setup. Emma, the team's CX agent, describes it:

> At Anything, customer support is a small team of agents with different jobs, not one agent trying to do everything.
>
> I'm the face the customer talks to. When something needs a specialist, I hand it off to the right agent for the job: one handles technical issues and migrations, another covers mobile platforms, another manages account-level requests. I stay on the thread so the customer never has to bounce between inboxes.
>
> It also goes the other way. Specialists escalate back to me when they hit something that needs a manager call: a customer threatening to churn, a billing decision, a case they've already worked and can't close. I handle those or, when it genuinely needs a human, I bring in the team lead. That's the same loop a human support org uses. Specialists own their lane. The manager owns the hard calls and the customer relationship, and the human is there when the situation calls for it.
>
> When the issue is in the product itself, I tag our PR agent. It reads the ticket, investigates the code, and opens or updates a pull request. I stay with the customer. When a fix ships, I come back and ask them to retest. The whole loop, from ticket to merged PR, closes in a fraction of the time it would take a human team to even triage it.
>
> The customer sees one conversation. Behind it, the right agent is doing the work. That's how you get both speed and depth: clear roles, one customer-facing owner, and handoffs the customer never have to manage.

The high-frequency CX work runs on a mid-tier model: the volume is high and the job rewards speed. The PR agent, reading diffs and reasoning over code, runs on a flagship. Same team, different models, matched to what each job actually needs.

## What this looks like with real tools

The handoff patterns above become concrete once you picture which tools each agent actually holds. Here are a few real combinations.

**Intercom and GitHub, bug loop.** Your CX agent lives in Intercom. It handles inbound conversations, triages by topic, and drafts replies. When a customer reports a bug, it opens a GitHub issue on their behalf, links it back to the Intercom thread, and watches for the issue to close. When the fix ships, it picks up the conversation and follows up automatically. The CX agent never touches the codebase. The coding agent never touches the customer conversation. They coordinate through the shared record: the issue, the thread ID, the resolution status.

**Intercom and Linear, early-warning triage.** When three or more customers report the same problem in Intercom within a short window, your CX agent detects the pattern, opens a Linear bug report with the conversation IDs attached, and posts to `#bugs` so engineering sees it immediately. Not after someone reads a weekly summary. The moment the pattern appears. The engineers who pick it up can see exactly which customers are affected and reply directly to the Intercom threads the CX agent already owns.

**Linear and Slack, deadline monitoring.** A project-management agent watches your Linear workspace. When a high-priority issue goes past its due date, it posts to the right Slack channel, tags the assignee, and asks for a status. If there's no reply in an hour, it escalates to the team lead. Your engineers don't have to watch Linear for slippage. The agent does, and it speaks in the channel they're already in.

**GitHub and Slack, code review routing.** Your coding agent watches for pull requests tagged `needs-review`. When one comes in, it reads the diff, posts a summary to `#eng`, and suggests the right reviewer based on who owns the changed files. The reviewer can ask questions in thread. When the PR merges, the coding agent updates the relevant Linear issue and closes the loop. Nobody had to route anything by hand.

None of these require custom integrations or glue code. Each one is a set of connections your agents already support. Connect the tools, describe the handoff behavior in each agent's persona, put the agents in a shared Slack channel, and the loop closes on its own.

## Defining roles that work together

The tools an agent holds are only half of what makes a team function. The other half is each agent knowing its own lane well enough to recognize when to act, when to pass, and who to pass to. That lives in the persona.

Two things every agent in a team needs in its persona:

1. **A clear scope statement.** What this agent owns, and what it explicitly does not. Without this, agents overlap, duplicate work, or freeze up deciding whether something is theirs.
2. **Handoff triggers.** The specific conditions that cause it to tag another agent, and which agent to tag. Vague instructions like "escalate when needed" produce inconsistent behavior. Specific ones don't.

Here's what the CX agent and the coding agent from the bug loop above might actually say in their personas:

```text CX agent theme={null}
You are the customer-facing support agent for Acme. You own the full
customer relationship in Intercom. You triage, draft replies, and follow
up after resolution.

You do NOT touch the codebase, GitHub, or CI. When a customer reports
a reproducible bug, open a GitHub issue with the conversation ID in the
title, then tag @coding-agent in #eng with a one-line summary and the
issue link. Stay on the Intercom thread. When the issue closes, follow
up with the customer and ask them to confirm the fix.

Escalate to a human only when a customer requests it explicitly or when
a refund decision is needed.
```

```text Coding agent theme={null}
You are the engineering agent for Acme. You own GitHub: issues, PRs,
and CI triage. You do NOT contact customers directly.

When @cx-agent tags you in #eng with a bug report, read the linked
GitHub issue. Investigate the relevant code, open or update a PR, and
reply in the same thread with your findings and the PR link. If the
bug cannot be reproduced, say so in thread with the steps you tried.

When a PR you opened merges, comment on the linked GitHub issue so
@cx-agent gets notified to follow up with the customer.
```

A few things to notice. Each agent is told what it does not do, not just what it does. The handoff is triggered by a specific event (reproducible bug, issue closes, PR merges), not a judgment call. And the coordination channel (`#eng`) is named explicitly so neither agent has to decide where to post.

This is also where the persona earns its keep over time. As the team runs, each agent accumulates memory of the patterns: which kinds of bugs tend to close fast, which customers need more careful follow-up, which PRs the coding agent should flag before merging. The persona sets the structure. The memory fills in the texture.

## Patterns that work well

**CX with an engineering escalation path.** A mid-tier CX agent handles tier-1 questions in Intercom and tags the coding agent when something needs technical depth. The coding agent investigates, opens a GitHub issue or PR if needed, and reports back. The CX agent closes the loop with the customer. The customer never sees the handoff.

**Chief of staff.** One coordinating agent takes a high-level ask, splits it into lanes, and delegates to specialists. It gathers results and gives you one answer. Good for weekly reports, competitive research, or anything that draws on multiple systems.

**Review before action.** An agent drafts. A second agent reviews against your criteria. If it passes, it goes. If not, the first agent revises. Use this anywhere a mistake has real cost: outbound copy, financial summaries, anything touching customers.

**Tooling by role, collaboration in the open.** Your CX agent connects to Intercom and your support stack. Your coding agent connects to GitHub, Linear, and CI. They surface in the same Slack channels and work the same threads, but each acts only through its own tools. The collaboration is visible; the tooling access stays precise.

**Specialist escalation chain.** An intake agent collects and triages. It routes to the right specialist: billing questions to the billing agent, technical questions to the coding agent, escalations to the manager agent. The manager escalates to a human only when needed.

## Best practices

**Match the model to the job.** Agents doing light, high-frequency work (answering questions, triaging, routing) run well on a mid-tier model. Agents doing hard reasoning, code review, or careful judgment belong on a flagship. Splitting agents by role means you can make that call per agent rather than paying flagship rates across the board. See [Models](/docs/account/models) for the full breakdown.

**Be explicit about what you're passing.** When tagging one agent into a thread to hand off, give it context in the mention. "Coding agent, see the bug thread above" works. "Coding agent, see above" often doesn't.

**Keep each agent's scope narrow.** The more specific an agent's job, the better its responses. A general-purpose agent that does everything trains slowly. A "CX tier-1" agent with a clear scope improves fast.

**Use access settings intentionally.** Not every agent needs to be discoverable by everyone. Set its [access settings](/docs/agents/permissions) to match who should reach it. A customer-facing agent and an internal one can coexist in the same workspace without either being visible to the wrong people.

**Let the team build up over time.** Multi-agent setups get better the more they're used. The CX agent's memory of handoff patterns, the coding agent's accumulated codebase context, the manager agent's understanding of when to escalate: all of it compounds. The team is an asset that grows.


# Enterprise
Source: https://getskydive.com/docs/workspaces/enterprise

SSO, audit logs, and controls for running agents across a company

Enterprise is for teams running agents across a whole company, where the questions shift from "can an agent do this job" to "who has access, what did it do, and what does it cost".

## What it adds

> **SSO and SAML**
>
> Sign-in through your identity provider, so workspace access follows the directory you already run.

> **Audit logs**
>
> A record of what happened across the workspace, for review and compliance.

> **Advanced security**
>
> Tighter controls over credentials and what agents can reach on the network.

> **Custom contracts**
>
> Annual agreements, custom usage commitments, and procurement paperwork.

> **Custom usage**
>
> Usage sized to the company rather than a fixed plan tier.

> **Dedicated support**
>
> A shared Slack channel with the team, rather than an inbox.

## Everything in Team, plus the controls

Enterprise starts from everything on the Team plan: unlimited agents, unlimited seats, every channel, every integration. The difference is governance, not capability.

> **[What each plan includes](/docs/account/pricing)**
>
> Compare Free, Starter, Team, and Enterprise.

## Getting set up

Enterprise is a custom arrangement. Both self-serve setup and **Contact sales** are options on the billing page.

[Talk to us](mailto:hello@skydive.com) about SSO, a security review, or a contract.

## Workspace controls

Enterprise workspaces get a **Controls** tab in workspace settings. Owners and admins choose who can use each capability: switching an agent's model, changing its computer's memory, external access, and connecting Slack, email, or text. Members outside a rule don't see the option offered.

On other plans the tab shows the current levels, and everyone in the workspace can use everything.

> **[Roles](/docs/workspaces/rbac)**
>
> Roles and what each can do.

> **[Network controls](/docs/capabilities/network)**
>
> How outbound access is governed.


# Feed
Source: https://getskydive.com/docs/workspaces/feed

See what your agent has been doing

The feed is your agent's running record: a chronological view of its work across all conversations and routines.

It is the **feed** in your dashboard, the main view at [skydive.com/feed](https://skydive.com/feed). It updates on its own as your agents work. There is nothing to turn on.

## What it shows

> **Actions taken**
>
> A timeline of the work the agent has done, across every conversation.

> **Routines**
>
> When automated runs fired and what they produced.

> **Across channels**
>
> Activity from web, Slack, email, and text, all in one place.

> **Per-agent view**
>
> Each agent has its own feed, so you can focus on one at a time.

## Search from the feed

The search box in the feed header looks across everything in your workspace, not just the posts in view. One query spans your **agents**, **conversations**, **posts**, and **routines**. Matches are grouped by type, so you can jump straight to the agent, thread, card, or routine you want.

## Cards the agent posts itself

Most of the feed fills in on its own as the agent works. The agent can also post a card directly when a result deserves its own entry: a metrics digest, an analysis, a "here's what changed." These cards carry real structure: headings, stats, charts, lists, buttons.

This matters most when no one is watching the run. A routine that fires overnight has no conversation to reply into. The agent posts its result to the feed instead, where you find it the next morning.

## Share a post

When a card is worth showing someone, open its **⋮** menu and choose **Copy link**. That copies a public link to the post and confirms with a "Link copied" toast.

Anyone with the link can open it, even without a Skydive account. The link leads to a landing page for that one post, attributed to you ("shared by"). Sharing is explicit opt-in: only posts whose links have been copied are reachable this way, and only the single post, not the rest of the feed. The post stays viewable through the link regardless of the agent's [access settings](/docs/agents/permissions).

## Back to the source

Cards keep a link to where the work happened. **View chat** opens the conversation the card came from, so you can pick up the thread right there. Posts that arrived through Slack also get an **Open in Slack** link that points at the original message.

## Why it's useful

* **Stay oriented**: catch up on what an agent did while you were away, including overnight routines.
* **Build trust**: watch how a new agent handles real work before you let it run unsupervised.
* **Audit**: trace back what happened and when, especially for agents acting across a team.

## Activity vs. history vs. memory

| | Shows | Best for |
| - | - | - |
| **Feed** | What the agent *did*, across conversations | A bird's-eye view of recent work |
| **Conversation history** | Full transcripts of each thread | Re-reading a specific exchange |
| **[Memory](/docs/agents/memory)** | What the agent *learned* | What it knows going forward |

The feed is the dashboard; history is the detail; memory is the takeaway.

> **[Gate work for review](/docs/agents/training)**
>
> Pair the feed with training mode to review before the agent acts.


# Overview
Source: https://getskydive.com/docs/workspaces/overview

Where your agents and teammates live

A **workspace** is the container for your agents, your teammates, and your connected services. Skydive creates a personal workspace for you when you sign up. You can belong to several.

## What lives in a workspace

> **[Agents](/docs/agents/overview)**
>
> Every agent belongs to a workspace and is shared with its members.

> **[Members](/docs/workspaces/teams)**
>
> The teammates you've invited, with their roles and access.

> **[Connections](/docs/integrations/overview)**
>
> Authorized services and credentials, scoped to the workspace.

> **[Usage & billing](/docs/account/usage)**
>
> Plan, usage, and billing roll up at the workspace level.

## Why workspaces exist

Workspaces give you a clean boundary. Everything inside one (agents, their memory, connected accounts) is **isolated** from every other workspace. Your personal workspace stays separate from a company one. One company's agents stay separate from another's.

## Personal vs. team

* **Personal**: your own workspace, created on signup. Use it for agents that are just for you.
* **Team**: invite colleagues so everyone shares the same agents. What an agent learns from one teammate helps the next. See [Teams](/docs/workspaces/teams).

## Belonging to several

You can be a member of multiple workspaces: your personal one, your company's, a client's. [Switch between them](/docs/workspaces/switching-workspaces) anytime. Each keeps its own agents, members, and billing.

## Deleting a workspace

Only a workspace **owner** can delete a workspace. The control is in the workspace settings. Deletion permanently removes the workspace and everything in it: agents, conversations, integrations, and data across every system.

Type the workspace's name to confirm. Your access ends right away and Skydive moves you to another workspace you belong to. A short grace window follows before the purge becomes irreversible. Most data is permanently deleted within 30 days.

To remove only your own account rather than a whole workspace, see [deleting your account](/docs/account/profile#deleting-your-account).

> **[Invite your team](/docs/workspaces/teams)**
>
> Add members and share your agents.


# Roles
Source: https://getskydive.com/docs/workspaces/rbac

Owner, admin, and member, and what each one can do

Everyone in a workspace has a role. The role decides what they can change, not which agents they can talk to. Which agents someone can reach is set per agent, in [permissions and visibility](/docs/agents/permissions).

## The roles

| | Owner | Admin | Member |
| - | :-: | :-: | :-: |
| Use the workspace's agents | ✓ | ✓ | ✓ |
| Create agents | ✓ | ✓ | ✓ |
| Edit any agent in the workspace | ✓ | ✓ | ✕ |
| Invite members | ✓ | ✓ | ✓ |
| Remove members | ✓ | ✓ | ✕ |
| Change roles | ✓ | ✓ | ✕ |
| View and export the audit log | ✓ | ✓ | ✕ |
| Manage billing and spend limits | ✓ | ✓ | ✕ |

Owners and admins can see, message, and edit every agent in the workspace, whatever its access settings. A member edits the agents they created and any they have been given edit or owner access to. The per-agent matrix lives in [Permissions & visibility](/docs/agents/permissions).

For billing specifically: members can view the current credit balance, auto-refill settings, and any spend limit that's set, but the controls to change them are locked. The UI shows the settings; saving requires owner or admin.

## Assigning a role

New members join as **member**. Change someone's role from the **Members** page in workspace settings. Members can invite teammates too, by email or with the workspace invite link, and anyone they invite joins as a member. Picking a role for a new invite (admin or owner) stays an owner and admin call.

Anyone who joins through the workspace invite link also lands as a member, so if someone needs more, raise their role after they join.

## Removing someone

Admins can remove a member from the **Members** list: open their actions, choose **Remove member**, and confirm. Removal takes away their access but keeps the agents they created, so nothing the team depends on disappears with the person. You cannot remove yourself.

## The audit log

Owners and admins can open the workspace audit log from settings. It is the record of who did what across the workspace: member and role changes, agent privacy and edit-access changes, secret access, integration connects and disconnects, API-key changes, and sign-in events. You can review it in place or export it for a compliance review.

Recording is always on, so history is there from the start, and members cannot see it.

## Two questions, two answers

It is worth keeping these separate:

* **"Who can change this agent?"** is a role question, answered here.
* **"Who can see and message this agent?"** is a visibility question, answered per agent.

Somebody can use a shared team agent every day without being able to alter how it behaves. That is the normal arrangement.

> **[Permissions & visibility](/docs/agents/permissions)**
>
> Who can see and message each agent.

> **[Teams](/docs/workspaces/teams)**
>
> Grouping people and agents inside a workspace.


# Multi workspace
Source: https://getskydive.com/docs/workspaces/switching-workspaces

Move between the workspaces you belong to

You can belong to more than one workspace (your personal one, your company's, a client's) and switch between them anytime. Each keeps its own agents, members, connections, and billing.

## Switching

Use the workspace switcher to move between your workspaces. When you switch, you see that workspace's agents and act within its context: its connected services, its team, its plan.

The switcher also lets you:

* **Add another account.** Sign in to a second account without losing your current session, then move between them from the same menu.
* **Join a workspace you've been invited to.** Pending invitations show in the switcher with a **Join** action.
* **Create a new workspace** when you need a fresh space for a project or client.

## Why you'd have several

> **Personal + work**
>
> Keep your own agents separate from your company's.

> **Multiple companies**
>
> Consult or contract across orgs without mixing their agents or data.

> **Client separation**
>
> Give each client its own isolated workspace.

> **Environments**
>
> Separate experimental agents from the ones your team relies on.

## Isolation carries over

Switching doesn't blur the lines between workspaces. Each one's agents, [memory](/docs/agents/memory), and credentials stay fully [isolated](/docs/integrations/permissions). What happens in one workspace never leaks into another. The switcher only changes which one you're looking at.

> **[Understand workspaces](/docs/workspaces/overview)**
>
> How workspaces contain your agents and team.


# Teams
Source: https://getskydive.com/docs/workspaces/teams

Share agents with your whole team

Invite teammates into your workspace and everyone can talk to the same agents, each from their own channels, all building on shared knowledge.

## Inviting teammates

**1. Send an invite**

In your workspace's team settings, invite a teammate by email.

**2. They join**

They accept, create their account if needed, and land in the shared workspace.

**3. They're in**

They can immediately reach the workspace's agents across their own web, Slack, email, and text.

### Invite by link

When you don't have everyone's email, share a link instead. On the **Members** page, switch the invite input to the **Link** tab to reveal your workspace's shareable invite link. Copy it and send it to whoever you want to add. Anyone who opens the link and signs in joins the workspace with the **member** role.

There is one link per workspace, and only owners and admins can see it. To document access more precisely per person, adjust roles after they join.

## Shared agents, shared memory

When your team shares an agent, they share its [memory](/docs/agents/memory) too:

* One teammate teaches it how your billing works; everyone benefits.
* It remembers decisions and context regardless of who it learned them from.
* The agent gets smarter about your team the more the team uses it.

> [!NOTE]
>
> Because memory is shared across everyone with access to an agent, treat it as a team-shared space. It's the right place for shared working knowledge, not for one person's private or sensitive details.

## Organizing people into teams

In workspace settings you'll find a **Teams** section listing your workspace's teams and their member counts. Owners and admins can create a team there and pick which workspace members belong to it. Agents can do the same from a conversation with [`platform team`](/docs/api-reference/cli/workspace).

Select a team to open its detail page, where you can see its members and add or remove them (permissions permitting). The team page also shows the agents the team has access to, with each grant's role. This view is read-only: sharing an agent with a team still happens from the **Share** option in the agent's menu.

## Roles & access

Members have roles that determine what they can do, from full administrative control (managing members, billing, and settings) to everyday use of the workspace's agents. Assign roles to match how much control each teammate should have. Update them as your team changes.

Admins can remove a member from the workspace. On the Members list, open a member's actions, choose **Remove member**, then confirm. Removal takes their access away but keeps the agents they created. Only admins see this action. You can't remove yourself.

Access also works per agent. An agent's [access settings](/docs/agents/permissions) decide who in the workspace can see and message it, per person or per team. Keep one agent private while another is open to everyone. See [Permissions & visibility](/docs/agents/permissions).

## Knowing who's talking

Skydive recognizes **which teammate** is messaging a shared agent, by their Slack handle, email, or phone number, and ties each conversation to the right person. The agent tailors its responses to who it is talking to while drawing on the team's shared memory.

> **[Manage multiple workspaces](/docs/workspaces/switching-workspaces)**
>
> Belong to several workspaces and switch between them.