> For the complete documentation index, see [llms.txt](https://docs.flatredball.com/gum/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.flatredball.com/gum/using-gum-with-ai/ai/mcp-server.md).

# MCP Documentation Server

Gum's documentation is available to AI assistants through a hosted **MCP server**. Connecting your assistant to it lets the assistant search and read the current Gum docs on demand, so its answers reflect the latest guidance instead of stale training data.

## What is MCP?

The **Model Context Protocol (MCP)** is an open standard that lets AI assistants connect to external tools and data sources. An MCP server exposes a set of capabilities — in this case, searching and fetching documentation pages — that a compatible assistant can call while it works.

Most modern AI coding tools support MCP, including Claude Code, Claude Desktop, Cursor, and VS Code.

## Endpoint

The Gum documentation MCP server is hosted at:

```
https://docs.flatredball.com/gum/~gitbook/mcp
```

It exposes tools to **search** the documentation and **fetch** individual pages, so your assistant can look up authoritative answers about layout units, Forms controls, code generation, the file formats, and everything else covered in these docs.

## Connecting your assistant

### Claude Code

Add the server from your terminal:

```
claude mcp add --transport http gum-docs https://docs.flatredball.com/gum/~gitbook/mcp
```

Once added, your assistant can query the Gum docs during any session. Use `claude mcp list` to confirm it is connected.

### Cursor, VS Code, Claude Desktop, and other clients

Most clients are configured with an `mcpServers` block in a JSON settings file. Add an entry pointing at the endpoint:

```json
{
  "mcpServers": {
    "gum-docs": {
      "url": "https://docs.flatredball.com/gum/~gitbook/mcp"
    }
  }
}
```

The exact file location and field names vary by client — check your tool's MCP documentation for where this config lives. Some clients label the remote-HTTP transport explicitly (for example with a `"type": "http"` field).

{% hint style="info" %}
A few clients can only launch MCP servers as local processes (stdio transport) and cannot connect to a remote URL directly. For those, bridge to the hosted server with [`mcp-remote`](https://www.npmjs.com/package/mcp-remote):

```json
{
  "mcpServers": {
    "gum-docs": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://docs.flatredball.com/gum/~gitbook/mcp"]
    }
  }
}
```

{% endhint %}

## Using it

Once connected, you do not need to do anything special — just ask your assistant Gum questions as usual. When it needs authoritative information, it can search and read the docs through the server rather than guessing. For tasks that involve writing or generating code, pair this with [GumCli for Agents](/gum/using-gum-with-ai/ai/gumcli-for-agents.md) so the assistant can also act on a project and verify the result.


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter:

```
GET https://docs.flatredball.com/gum/using-gum-with-ai/ai/mcp-server.md?ask=<question>&goal=<endgoal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal.

The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.