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

# MCP server

> Give your AI coding assistant live access to Torpedo's documentation

Torpedo runs a public [Model Context Protocol](https://modelcontextprotocol.io) server so AI coding assistants (Claude, Cursor, and other MCP clients) can search and read Torpedo's documentation directly instead of relying on training data that may be outdated.

<Info>
  This server is documentation-only. It never sends email or SMS, proxies Torpedo API calls, or accesses your account. Never paste a Torpedo API key into an MCP tool call — application code should read `TORPEDO_API_KEY` from server-side secret storage.
</Info>

## Endpoint

```
https://mcp.torpedo.co.mz/mcp
```

The server is stateless, requires no authentication, and speaks streamable HTTP JSON-RPC. Requests are capped at 64 KiB and rate-limited to 60 requests per minute per IP.

## Connect a client

<Tabs>
  <Tab title="Claude Code">
    ```bash theme={null}
    claude mcp add --transport http torpedo-docs https://mcp.torpedo.co.mz/mcp
    ```
  </Tab>

  <Tab title="Claude Desktop / claude.ai">
    Add a custom connector under **Settings → Connectors** with the URL above.
  </Tab>

  <Tab title="Cursor">
    Add to `.cursor/mcp.json`:

    ```json theme={null}
    {
      "mcpServers": {
        "torpedo-docs": {
          "url": "https://mcp.torpedo.co.mz/mcp"
        }
      }
    }
    ```
  </Tab>
</Tabs>

## Tools

| Tool                        | Description                                                                                                                   |
| --------------------------- | ----------------------------------------------------------------------------------------------------------------------------- |
| `search_torpedo_docs`       | Search documentation pages and API operations by keyword. Accepts an optional `section` filter and a result `limit` (max 10). |
| `get_torpedo_doc`           | Retrieve one documentation page in full by `slug`.                                                                            |
| `get_torpedo_api_operation` | Retrieve a structured OpenAPI operation by `operationId`, including auth, parameters, schemas, responses, and related guides. |

Every page and API operation is also exposed as an MCP resource, so clients that support resource browsing can list and read them without calling a tool first.

All source content is the same English-language documentation and OpenAPI spec published on this site, kept in sync automatically — there's no separate copy to fall out of date. Queries in other languages (e.g. Portuguese) are accepted, but returned content is always the canonical English source.