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

# Brimble MCP

> The official Model Context Protocol server for Brimble. Connect Claude, Cursor, and other agents so they can manage projects, sandboxes, domains, and more on your account.

**Brimble MCP** is Brimble's official [Model Context Protocol](https://modelcontextprotocol.io) server at **`https://mcp.brimble.io`**. It exposes Brimble platform tools (projects, sandboxes, logs, domains, and more) so AI clients can act on your account with your permission, using an API key you control.

## What you can do with it

Depending on the permissions on your API key, agents can:

* List and inspect **projects** and deployments
* Work with **sandboxes**, volumes, and related tooling
* Read **logs**, metrics-style account views, and activity
* Manage **domains**, networking, and other platform surfaces exposed as tools

Tool availability follows the **scopes on the API key**. A read-only key cannot deploy; a sandbox-only key cannot manage domains. Create a scoped key in the dashboard when you want least privilege.

## How it works

1. You create a Brimble **API key** in the dashboard (see [API keys](/security/api-keys)).
2. You configure your agent to call **`https://mcp.brimble.io`** with that key.
3. The agent can use tools within that key's permissions on your account.

Optional: a **connect** flow in the dashboard can verify that a pasted key belongs to the signed-in user and show its permissions before you paste the key into an agent.

## Authentication

Brimble MCP accepts either:

```http theme={null}
x-brimble-key: br_...
```

or:

```http theme={null}
Authorization: Bearer br_...
```

Both mean the same thing: your **platform API key**.

Optional team context:

```http theme={null}
x-brimble-team: <team-id>
```

<Warning>
  Treat the API key like a password. Anyone who has it can use every tool your key allows, including write and spend-related actions if those scopes are granted. Prefer scoped keys and non-production keys when testing agents.
</Warning>

## Endpoints

| URL                             | Role                         |
| ------------------------------- | ---------------------------- |
| `https://mcp.brimble.io`        | MCP endpoint (agent traffic) |
| `https://mcp.brimble.io/health` | Liveness check               |

`GET /` without an MCP session returns an error by design. Use `/health` to check the service, not the root path in a browser.

## Security notes

* **Env vars and secrets:** Prefer setting secrets in the dashboard; use the agent for redeploys and non-secret configuration.
* **2FA-gated actions:** Some account actions require two-factor authentication in the dashboard. MCP will not complete 2FA for you; finish those steps in the UI.
* **Revocation:** Reset or revoke the API key in the dashboard to cut off all agents using that key immediately.

## Next steps

* [Connect Claude Code, Cursor, or a remote client →](/mcp/connect)
* [Create and manage API keys →](/security/api-keys)