Docmost MCP Server

An MCP server that gives AI assistants direct access to your self-hosted Docmost documentation via its API.

Works with any MCP-compatible client, including:

• Claude Code / Claude Desktop

• Cursor

• Windsurf

• VS Code (GitHub Copilot)

• Cline

• Continue

• Zed

Features

Prerequisites

• Python 3.10+

• A running Docmost instance with valid user credentials

• An MCP-compatible client (Claude Code, Claude Desktop, Cursor, VS Code, etc.)

Installation

1. Clone the repository

2. Create a virtual environment and install dependencies

This installs two dependencies:

• mcp — official Python SDK for building MCP servers

• requests — HTTP client for Docmost API calls

3. Configure credentials

Copy the example config and fill in your details:

Edit config.json:

Note: config.json contains sensitive credentials and is excluded from version control via .gitignore.

4. Verify the setup

Expected output:

If you see your spaces and search results, the client is working correctly.

Connecting to an MCP client

The server uses stdio transport, which is supported by all major MCP clients. Below are setup instructions for the most popular ones.

Claude Code

Choose one of the following options.

Important: Use absolute paths to both the Python binary inside venv and mcp_server.py.

Option 1: Global config (recommended)

Add to ~/.claude/settings.json:

Option 2: Project-level config

Create .claude/settings.json in your project root:

Option 3: CLI command

After adding the config, restart Claude Code or start a new session.

Claude Desktop

Add the server to your Claude Desktop config file:

• Linux: ~/.config/claude/claude_desktop_config.json

• macOS: ~/Library/Application Support/Claude/claude_desktop_config.json

• Windows: %APPDATA%\Claude\claude_desktop_config.json

Restart Claude Desktop after saving the config.

Cursor / Windsurf / VS Code / Other clients

Most MCP clients use the same configuration format. Add the server with:

• Command: /absolute/path/to/docmost-mcp/venv/bin/python

• Args: /absolute/path/to/docmost-mcp/mcp_server.py

• Transport: stdio

Refer to your client's documentation for the exact config file location and format.

Usage

Once the MCP server is connected, your AI assistant can use the following tools:

List all spaces

Search documentation

Search within a specific space

Get a specific page

How it works

1. The MCP client launches the server as a subprocess and communicates via stdio (stdin/stdout).

2. On the first API call, the server authenticates with your Docmost instance using email/password and receives a JWT token from the Set-Cookie header.

3. The token is cached in token.json for subsequent requests. If a request returns 401, the server automatically re-authenticates and retries.

4. Page content is stored in Docmost as ProseMirror JSON. The server converts it to clean Markdown (headings, lists, tables, code blocks, images, links, etc.) before returning it to the client.

Project structure

Security

• config.json contains your credentials — never commit it to git

• token.json contains a JWT token — never commit it to git

• Both files are already listed in .gitignore

Troubleshooting

"No authToken in response cookies"

• Verify that email and password in config.json are correct

• Verify that base_url points to your Docmost instance

"Connection refused"

• Check that your Docmost instance is running and accessible

• Check your network connection and firewall rules

"401 Unauthorized"

• The token may have expired — delete token.json and it will be recreated automatically on the next request

MCP tools not appearing in the client

• Ensure you used absolute paths to venv/bin/python and mcp_server.py

• Check that the virtual environment has all dependencies installed (pip install -r requirements.txt)

• Restart your MCP client after changing the config

License

MIT