Documentation Search MCP Server

Documentation Search MCP Server

A Model Context Protocol (MCP) server that provides intelligent documentation search capabilities across multiple programming libraries and frameworks.

Features

• Multi-library support: Search documentation for 30+ popular libraries and frameworks
• External configuration: Easily add new documentation sources without code changes
• Web search integration: Uses Serper API for targeted documentation searches
• Content extraction: Fetches and parses documentation content from official sources
• Smart caching: In-memory caching with TTL to improve response times and reduce API calls
• Enhanced reliability: Exponential backoff retry logic for robust error handling
• Auto-suggestions: Library name auto-completion for better user experience
• Health monitoring: Built-in health checks for documentation sources
• Concurrent processing: Parallel fetching of multiple documentation pages

Supported Libraries

The server currently supports documentation search for:

• Programming Languages: Python, JavaScript, SQL
• Web Frameworks: React, Node.js, Express, Django, Flask, FastAPI
• Databases: MongoDB, SQLAlchemy, PostgreSQL
• AI/ML: OpenAI, Anthropic, LangChain, LlamaIndex, LangGraph
• Cloud & DevOps: AWS, Kubernetes, Docker, Terraform, Ansible
• Data Science: Pandas, NumPy, Matplotlib
• Development Tools: Git, Streamlit, Supabase, Firebase, Vercel

And many more! See config.json for the complete list.

Setup

Prerequisites

• Python 3.8+
• UV package manager (recommended) or pip

Installation

1. Clone this repository:

2. Install dependencies:

3. Set up your environment variables:

4. Get a Serper API key:
   • Visit serper.dev
   • Sign up for a free account
   • Copy your API key to the .env file

Configuration

Adding New Documentation Sources

To add support for new libraries, simply edit the config.json file:

The server will automatically pick up changes to this file on restart. No code modifications required!

Configuration Structure

The config.json file contains the following sections:

Cache Configuration

• enabled: Whether to enable in-memory caching
• ttl_hours: Time-to-live for cached content (default: 24 hours)
• max_entries: Maximum number of cached entries (default: 1000)

Documentation URLs

• Key: Library/framework name (used in search queries)
• Value: Base URL of the official documentation

Usage

Running the Server

The server runs using the stdio transport protocol for MCP communication.

Integration with AI Tools

Adding to Cursor

1. Open Cursor Settings (Cmd/Ctrl + ,)
2. Navigate to "Features" → "Model Context Protocol"
3. Add a new MCP server configuration:

4. Replace /path/to/your/documentation/main.py with the absolute path to your main.py file
5. Save the configuration and restart Cursor

Adding to Claude Desktop

1. Locate your Claude Desktop configuration file:
   • macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
   • Windows: %APPDATA%\Claude\claude_desktop_config.json
2. Add the MCP server configuration:

3. Replace /path/to/your/documentation/main.py with the absolute path to your main.py file
4. Restart Claude Desktop

Using Virtual Environments

If you're using a virtual environment (recommended), use the full path to the Python executable:

For UV:

For Conda:

Available Tools

The server provides several MCP tools for documentation search and management:

1. get_docs - Main Documentation Search

Parameters:

• query (string): Your search query (e.g., "authentication", "database connection")
• library (string): The library to search in (e.g., "fastapi", "react", "python")

Example:

2. suggest_libraries - Library Auto-Completion

Parameters:

• partial_name (string): Partial library name for suggestions (e.g., "lang" → ["langchain"])

Example:

3. health_check - Documentation Source Health

Monitor the availability and response times of documentation sources.

Example:

4. get_cache_stats - Cache Statistics

View current cache usage and performance metrics.

Example:

5. clear_cache - Cache Management

Force fresh fetches by clearing the documentation cache.

Example:

Project Structure

How It Works

1. Query Processing: Takes your search query and target library
2. Cache Check: Checks in-memory cache for previously fetched content
3. Site-specific Search: Uses Serper API to search within the specific documentation site
4. Parallel Fetching: Concurrently fetches multiple documentation pages
5. Content Extraction: Parses and extracts clean text content using BeautifulSoup
6. Intelligent Caching: Stores results with TTL for faster future requests
7. Error Recovery: Implements retry logic with exponential backoff for reliability

Performance Features

• Smart Caching: Reduces API calls and improves response times by up to 10x for repeated queries
• Concurrent Processing: Fetches multiple documentation pages simultaneously
• Retry Logic: Exponential backoff ensures reliable operation even with network issues
• Content Optimization: Removes navigation, scripts, and styling for cleaner text extraction
• Memory Management: Automatic cache cleanup prevents memory bloat

API Dependencies

• Serper API: For web search functionality
• FastMCP: MCP server framework
• httpx: Async HTTP client
• BeautifulSoup4: HTML parsing and text extraction

Environment Variables

Create a .env file with:

Contributing

To add support for new libraries:

1. Add the library and its documentation URL to config.json
2. Test that the documentation site returns useful content
3. Submit a pull request

Troubleshooting

Common Issues

• "Library not supported": Check that the library name matches an entry in config.json
• "No results found": The search query might be too specific, or the documentation site might not be indexed well
• Timeout errors: Some documentation sites may be slow to respond; this is handled gracefully

AI Tool Integration Issues

• Tool not appearing: Ensure the absolute paths are correct and the Python environment has all dependencies installed
• Permission errors: Make sure the Python executable and main.py file have proper permissions
• Environment variables: Verify that SERPER_API_KEY is correctly set in the MCP configuration, not just your shell environment
• Virtual environment: Double-check you're using the correct Python path from your virtual environment

Debugging

The server includes error handling for:

• Network timeouts
• Invalid library names
• Empty search results
• Malformed configuration files

License

This project is open source. See LICENSE file for details.

documentation-mcp

documentation-mcp

documentation-mcp