Docs MCP Server · DeutschlandGPT

DeutschlandGPT exposes its entire product documentation, including the full OpenAI-compatible Platform API reference, through a public Model Context Protocol (MCP) server. Point an MCP-capable assistant such as Claude or Cursor at it, and the assistant can list, search, and read every documentation page on demand instead of relying on stale training data.

Endpoint

CODE

https://dialog.deutschlandgpt.de/docs/mcp

The server is public and read-only. No API key or authentication is required, and it only ever returns documentation that is already published here.

It speaks the Streamable HTTP transport and runs statelessly (a fresh server instance per request, no session state), so any standard MCP client can connect over plain HTTP.

Connecting a client

Add the server to your MCP configuration:

JSON

{
  "mcpServers": {
    "deutschlandgpt-docs": {
      "type": "http",
      "url": "https://dialog.deutschlandgpt.de/docs/mcp"
    }
  }
}

Restart the client and the list_docs, search_docs, and read_doc tools become available.

Available tools

The server registers three tools. Every tool accepts an optional language argument ("en" or "de", defaulting to "en").

list_docs

Enumerate every documentation page with its title, description, URL path, and slug array.

search_docs

Case-insensitive keyword search across titles, descriptions, and Markdown bodies. Returns ranked matches with a short snippet around each hit.

read_doc

Return the full raw Markdown of a single page. Accepts a doc URL or a slug path.

`list_docs`

Argument	Type	Description
`language`	`"en"` \| `"de"`	Optional. Documentation language.

Returns the full page index. Use it as a first call to discover what exists.

`search_docs`

Argument	Type	Description
`query`	`string`	Required. Keyword or phrase to search for.
`language`	`"en"` \| `"de"`	Optional. Documentation language.

Matches in titles are weighted highest, then descriptions, then body text.

`read_doc`

Argument	Type	Description
`path`	`string`	Required. A doc URL (`/docs/en/platform-api/quickstart`) or slug path (`platform-api/quickstart`).
`language`	`"en"` \| `"de"`	Optional. Documentation language.

Languages

All documentation is available in English (en, the default) and German (de). Pass the language argument to any tool to switch.

Typical flow

Call list_docs (or search_docs with a keyword) to find the relevant page.

Take the url or slugs from the result and pass it to read_doc.

The assistant now has the page's full Markdown source as grounding context.

Looking for the API itself rather than the docs? See the Platform API section for the OpenAI-compatible chat, embeddings, image, video, and audio endpoints.