This MCP server exposes Linux man pages via the MCP interface, enabling AI assistants to search, retrieve, and browse documentation from your local system. Core tools include search_man_pages for keyword or topic queries, get_man_page for full page content (with optional section), and list_man_sections to view available sections (1–9) with descriptions. Pages are exposed as resources using man:// URIs, such as man://search/network for search results or man://1/ls for the ls page in section 1. To use it in an environment like Claude or GitHub Copilot, configure the MCP client to start the server and then query via the provided tools to fetch and format man pages for AI consumption.

Prerequisites:

• Linux with standard man pages and apropos installed
• Python 3.10 or higher
• MCP library (bundled in MCPB or installable via package managers)
• Optional: uv (Hyperfast Python runner) for development workflows

Install steps (development workflow via uv):

1. Clone the repository git clone https://github.com/guyru/man-mcp.git cd man-mcp

2. Install dependencies and run the server (development) uv sync

3. Run the server (development) uv run python3 server/main.py

   Alternatively, run directly with Python

   python3 server/main.py

4. If you plan to use MCP development tools uv run mcp dev server/main.py

Install steps (manual Python package install):

1. Clone the repository git clone https://github.com/guyru/man-mcp.git cd man-mcp

2. Install the package and dependencies from pyproject.toml pip install .

   Or editable development install

   pip install -e .

   For development with optional dependencies

   pip install -e .[dev]

3. Run the server (development) uv run python3 server/main.py

   or

   python3 server/main.py

Tips and notes:

• The server relies on the system's man pages; ensure your Linux environment has man and apropos available.
• If apropos retrieval fails, the server can fall back to man -k to locate relevant pages.
• Content is cleaned/formatted for AI-friendly consumption and includes timeout protections for asynchronous operations.
• Environment variables like PYTHONPATH may be needed if running via direct Python paths; refer to the VS Code Copilot guidance in the README for configuring paths.
• MCP bundles (MCPB) are supported for some clients; for others, use the explicit uv or python startup commands shown in the installation section.
• When using the MCP bundle, the bundle file must be loaded by your client (e.g., Claude Desktop or VS Code integrations) as described.