DocsetMCP

DocsetMCP is a Model Context Protocol (MCP) server that lets you query and browse local Dash docsets from within AI assistants and MCP clients. It provides fast, offline access to documentation and cheatsheets by searching across installed docsets, filtering by language, and drilling into framework or class entries to reveal relevant notes. The server is designed to work with MCP clients like Claude Desktop, Cursor, VS Code, and other MCP-compatible tools, enabling seamless, private documentation retrieval without leaving your conversation. Available features include multi-docset search, language filtering, name-based matching, smart ranking, and container or framework guidance for exploring members. In addition, you can access cheatsheets (e.g., Git, Vim, Docker) and browse by category, with the option to search within each cheatsheet.

To use DocsetMCP, configure your MCP client to launch the server via uvx with the appropriate subcommand. For example, the Quick Start section demonstrates using uvx with the docsetmcp command. Once configured, you can ask natural-language questions like “Find me the AppIntent documentation” and receive targeted results from your local docsets, including exact, prefix, and substring matches prioritized by relevance.

The server is also designed to work with additional search paths and custom docset locations through environment variables and CLI options. This makes it easy to include third-party or shared docsets and integrate DocsetMCP into existing development environments, CI workflows, or personal toolchains.

Prerequisites

• macOS (Dash is Mac-only) or a development environment with a compatible MCP client
• Dash app with desired docsets downloaded
• Python 3.10 or higher
• UV package manager (for uvx-based usage)
• An MCP-compatible client (e.g., Claude Desktop, Claude Code CLI, Cursor, VS Code)

Installation options

1. No installation required (recommended if your MCP client supports uvx)

• If your MCP client can run uvx-based commands, you can use DocsetMCP without installing a package locally. Example configuration:

{
  "mcpServers": {
    "docsetmcp": {
      "command": "uvx",
      "args": ["docsetmcp"]
    }
  }
}

Note: uvx will download and run the package as needed when invoked by your MCP client.

2. Manual installation (local Python package)

• Install locally using pip and run the local entry point instead of uvx:

pip install docsetmcp

• Use docsetmcp in your MCP configuration instead of uvx docsetmcp:

{
  "mcpServers": {
    "docsetmcp": {
      "command": "python",
      "args": ["-m", "docsetmcp"]
    }
  }
}

3. Development installation

• Clone the repository and install in editable mode:

git clone https://github.com/codybrom/docsetmcp.git
cd docsetmcp
pip install -e .

• Run tests (optional):

pip install pytest pytest-cov pytest-xdist
pytest tests/test_docsets.py::TestDocsets::test_yaml_structure -v

Tips and common issues:

• Ensure Dash is installed and the desired docsets are downloaded in Dash.
• If you use custom docset or cheatsheet locations, set DOCSET_PATH and CHEATSHEET_PATH in your environment or pass appropriate CLI arguments.
• Use the list_available_docsets command to verify what docsets are detected on your system.
• For MCP client configuration, the uvx-based workflow relies on the client supporting uvx commands or equivalent wrappers.
• If you encounter connectivity or search issues, verify that the local docset database is accessible and that the uvx/docsetmcp executable is in your PATH.
• You can combine multiple search paths via additional CLI options as described in the configuration guidance to include third-party docsets.
• The project exposes a Python package (docsetmcp); if you’re integrating into a larger Python workflow, you can import modules and run tests directly.