> ## Documentation Index
> Fetch the complete documentation index at: https://getskydive.com/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# 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

<Tabs>
  <Tab title="MCP server">
    Connect Replicate over MCP for direct access to its tools.

    <Steps>
      <Step title="Add the connection">
        From the agent's tools, add Replicate's MCP server.
      </Step>

      <Step title="Authorize">
        A connect card appears. Authorize and approve access.
      </Step>

      <Step title="The agent gets to work">
        The agent can now reach Replicate through the MCP server.
      </Step>
    </Steps>
  </Tab>

  <Tab title="API key">
    Connect Replicate with an API key.

    <Steps>
      <Step title="Add the connection">
        Generate an API key from your Replicate account settings.
      </Step>

      <Step title="Authorize">
        Copy the key. It is only shown once.
      </Step>

      <Step title="The agent gets to work">
        Paste it when prompted in the agent's connection card for Replicate.
      </Step>
    </Steps>
  </Tab>
</Tabs>

## 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.

<CardGroup cols={2}>
  <Card title="How credentials stay safe" icon="shield-halved" href="/docs/integrations/permissions">
    The proxy model that keeps tokens out of the model's reach.
  </Card>
</CardGroup>