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

# MCP server

> Let AI agents like Claude and ChatGPT run Streamloop as tools — create, control, and monitor streams over the Model Context Protocol.

Streamloop runs a hosted **Model Context Protocol (MCP)** server so AI agents can manage streams as tools — uploading media, building playlists, connecting destinations, and starting, scheduling, and monitoring 24/7 streams. It's a thin, OAuth-protected layer over the same API as everything else, so an agent can do what you can do in the dashboard.

## Endpoint

|           |                                  |
| --------- | -------------------------------- |
| URL       | `https://mcp.streamloop.app/mcp` |
| Transport | Streamable HTTP                  |
| Auth      | OAuth 2.1 (PKCE, S256)           |

## Connect an agent

Add the server URL to any MCP-capable client. On first use the client opens Streamloop's OAuth flow; you sign in and approve the scopes, and the agent gets a token — no API key to copy or paste.

<Tabs>
  <Tab title="Claude">
    In Claude (Settings → Connectors / "Add custom connector"), add an MCP server with the URL `https://mcp.streamloop.app/mcp`, then complete the sign-in prompt.
  </Tab>

  <Tab title="ChatGPT">
    Add a connector / custom MCP server pointing at `https://mcp.streamloop.app/mcp` and authorize when prompted.
  </Tab>

  <Tab title="Custom client">
    Point any MCP client that speaks **Streamable HTTP** at `https://mcp.streamloop.app/mcp`. The server advertises OAuth via RFC 9728 protected-resource metadata and responds with a `WWW-Authenticate` challenge, so standards-compliant clients discover the authorization server automatically. See [Authentication](/api-reference/authentication).
  </Tab>
</Tabs>

## Permissions

The agent only gets the scopes you approve during sign-in. Grant the least it needs:

| Scope                    | Lets the agent                                            |
| ------------------------ | --------------------------------------------------------- |
| `streamloop:read`        | Read streams, destinations, playlists, and billing status |
| `streamloop:write`       | Create and control streams, destinations, and playlists   |
| `streamloop:destructive` | Delete streams and destinations                           |
| `streamloop:billing`     | Read billing and usage, and create top-up links           |

## Built-in safety

* **Destructive and financial tools are marked.** Tools that delete resources or spend money carry a `destructiveHint` and require explicit confirmation (`confirm: true`) before they run, so an agent can't quietly delete a stream or create a top-up by accident.
* **Scope-gated.** Each tool requires a specific scope; a tool the token isn't scoped for isn't callable.

<Note>
  The MCP tools mirror the REST API one-to-one — same operations, same data. See the full list on the [Tools](/api-reference/mcp/tools) page.
</Note>