What is a Skydive agent, exactly?
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?
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?
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?
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, so you can read, edit, or delete entries directly. See Memory.
Is my data and are my credentials safe?
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.
Can my whole team use the same agent?
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.
What channels can I reach my agent on?
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.
Can an agent run on its own schedule?
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.
How am I billed?
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 dashboard. See Pricing.
Can I run more than one agent?
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 and Pricing.
What happens to an agent's work when its sandbox restarts?
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 and survive restarts.