MCP Documentation Search Server

What is MCP Documentation Search Server?

This server enables AI systems to intelligently search across multiple popular framework and library documentations using a unified interface. Built with FastMCP, it provides a single tool to query documentation from sources like LangChain, Next.js, Tailwind CSS, Framer Motion, and more.

How to use MCP Documentation Search Server?

Clone the repository, set up a Python virtual environment, install dependencies with pip install -r requirements.txt, and run python main.py. Then use the get_docs function, passing a query and a library name (e.g., get_docs(query="animations", library="framer-motion")). The server intelligently handles multiple library name variations.

Key features of MCP Documentation Search Server

• Multi-library support across six popular frameworks/libraries
• Intelligent search with smart name resolution for library variations
• DuckDuckGo-powered search for accurate results
• Asynchronous processing and parallel content fetching
• Robust error handling and network timeout management
• Comprehensive test suite for reliability

Use cases of MCP Documentation Search Server

• AI assistants retrieving documentation for user queries without manual browsing
• Developers integrating doc search into automated workflows or coding tools
• Rapid prototyping by searching across multiple frameworks from a single interface
• Serving as a backend for documentation bots or chat interfaces

FAQ from MCP Documentation Search Server

What are the runtime requirements?

Python 3.8+ and either pip or uv package manager. A virtual environment is recommended.

How do I add a new documentation source?

Add the URL to DOCS_URLS in config.py and add common aliases to LIBRARY_ALIASES. Then update documentation and submit a pull request.

What should I do if I get a TimeoutError or no results?

Increase HTTP_TIMEOUT in config.py (default 30 seconds) for timeouts. For no results, try different search terms or verify the library name. Check your internet connection and the documentation URL for HTTP errors.

How does library name resolution work?

The system accepts multiple variations of library names (e.g., "framer", "framermotion", "framer-motion", "motion") and maps them to the correct documentation URL using aliases defined in config.py.

Are there any tests?

Yes, the project includes unit tests (for utilities and services), integration tests (for the main API), and uses pytest. Run all tests with python -m pytest.