abhiiiman/Web-Docs-MCP-Server

📚 WebDocs MCP Server

AI-Powered Documentation Search & Retrieval System
A Model Context Protocol (MCP) server that fetches real-time documentation from popular libraries and frameworks using web search.

🌟 Features

• 🔍 Real-time Documentation Search - Fetches the latest docs from official sources
• 🚀 Multiple Library Support - Langchain, Llama-Index, OpenAI, UV, and Qubrid
• 🧹 Clean Content Extraction - Removes HTML noise and returns readable text
• 🔗 Source Attribution - Every piece of content includes its source URL
• ⚡ Fast & Async - Built with async/await for optimal performance
• 🤖 MCP Compatible - Works seamlessly with Claude Desktop and other MCP clients

Claude Desktop Usage Demo Screenshots

📦 Supported Libraries

🛠️ Installation

Prerequisites

• Python 3.10 or higher
• UV package manager

Setup

1. Clone the repository

   git clone <your-repo-url>
   cd Docu_MCP
   

2. Install dependencies

   uv sync
   

3. Set up environment variables

   Create a .env file in the project root:

   SERPER_API_KEY=your_serper_api_key_here
   GROQ_API_KEY=your_groq_api_key_here
   

🔑 API Keys Setup

1. Serper API Key (Required)

The Serper API is used for web search functionality.

1. Visit serper.dev
2. Sign up for a free account
3. Navigate to your dashboard
4. Copy your API key
5. Add it to your .env file

2. Groq API Key (Optional)

Currently reserved for future LLM integration features.

1. Visit console.groq.com
2. Sign up and get your API key
3. Add it to your .env file

🚀 Usage

Option 1: Using the MCP Client (Python)

Test the server directly with the included client:

uv run mcp_client.py

Example output:

Available Tools: ['get_docs']
[Documentation content with sources...]

You can customize the query in mcp_client.py:

query = "How to setup ComfyUI AI ML Template?"
library = "qubrid"
res = await session.call_tool(
    "get_docs",
    arguments={"user_query": query, "library": library}
)

Option 2: Using Claude Desktop

Step 1: Configure Claude Desktop

Open your Claude Desktop config file:

• Windows: %APPDATA%\Claude\claude_desktop_config.json
• macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
• Linux: ~/.config/Claude/claude_desktop_config.json

Add the WebDocs server configuration:

{
  "mcpServers": {
    "WebDocs": {
      "command": "C:\\Users\\YOUR_USERNAME\\.local\\bin\\uv.EXE",
      "args": ["--directory", "D:\\Path\\To\\Docu_MCP", "run", "mcp_server.py"],
      "env": {
        "SERPER_API_KEY": "your_serper_api_key_here",
        "GROQ_API_KEY": "your_groq_api_key_here"
      }
    }
  }
}

⚠️ Important: Replace YOUR_USERNAME and D:\\Path\\To\\Docu_MCP with your actual paths!

Step 2: Restart Claude Desktop

1. Completely close Claude Desktop
2. Reopen Claude Desktop
3. The WebDocs tool should now be available

Step 3: Use the Tool

In Claude Desktop, you can ask:

"Use the WebDocs tool to search for how to create a ReAct agent in Langchain"

Claude will automatically call the get_docs tool with the appropriate parameters.

Option 3: Using MCP Inspector (Debugging)

The MCP Inspector is a powerful tool for testing and debugging your MCP server.

Start the Inspector

npx @modelcontextprotocol/inspector uv --directory "D:\AbhiiiMan Codes\Docu_MCP" run mcp_server.py

This will:

1. Start the MCP server
2. Launch a web interface at http://localhost:5173
3. Allow you to interactively test the server

Using the Inspector

1. Open your browser to the provided URL
2. Click "Connect" to establish connection
3. Navigate to "Tools" tab
4. Select get_docs tool
5. Fill in the parameters:
   • user_query: Your search query
   • library: One of langchain, llama-index, openai, uv, or qubrid
6. Click "Run Tool" to test

Troubleshooting

If port is already in use:

# Find and kill the process using port 6277
netstat -ano | findstr :6277
taskkill /PID <process_id> /F

🔧 Available Tools

get_docs

Searches and retrieves documentation content from supported libraries.

Parameters:

• user_query (string, required): The search query
  • Example: "How to use Langchain with OpenAI"
• library (string, required): The library to search
  • Options: langchain, llama-index, openai, uv, qubrid

Returns:

• Text content from documentation with source URLs

Example:

await session.call_tool(
    "get_docs",
    arguments={
        "user_query": "vector store integration",
        "library": "langchain"
    }
)

📁 Project Structure

Docu_MCP/
├── 📄 mcp_server.py          # Main MCP server implementation
├── 📄 mcp_client.py          # Test client for the server
├── 📄 constants.py           # Configuration and constants
├── 📄 utils.py               # Utility functions (HTML cleaning)
├── 📄 test_server.py         # Server launch test script
├── 📄 .env                   # Environment variables (create this)
├── 📄 pyproject.toml         # Project dependencies
└── 📄 README.md              # This file

🔍 How It Works

graph LR
    A[User Query] --> B[MCP Server]
    B --> C[Serper API]
    C --> D[Google Search]
    D --> E[Documentation URLs]
    E --> F[Fetch Content]
    F --> G[Clean HTML]
    G --> H[Return Text]
    H --> I[User/Claude]

1. Query Construction: Combines library domain with user query
2. Web Search: Uses Serper API to search Google
3. Content Fetching: Retrieves raw HTML from documentation pages
4. Content Cleaning: Extracts readable text using Trafilatura
5. Response Formation: Formats content with source attribution

🐛 Debugging & Testing

Test Server Launch

Run the test script to verify configuration:

uv run test_server.py

Expected output:

✅ Server process started successfully!
✅ Server is running and accepting connections!

Common Issues

Enable Debug Logging

Add to your .env:

LOG_LEVEL=DEBUG

🎯 Example Use Cases

1. Learning New Framework

Query: "Getting started with vector stores"
Library: "langchain"
→ Returns: Setup guides, installation steps, basic examples

2. Troubleshooting

Query: "Error handling in async queries"
Library: "llama-index"
→ Returns: Error handling patterns, best practices

3. API Reference

Query: "Chat completion parameters"
Library: "openai"
→ Returns: Parameter documentation, examples, limits

4. Tool Setup

Query: "Installing UV on Windows"
Library: "uv"
→ Returns: Installation guide, configuration steps

🤝 Contributing

Contributions are welcome! Here's how you can help:

1. Add More Libraries: Update constants.py with new documentation sources
2. Improve Content Cleaning: Enhance the HTML extraction in utils.py
3. Add Features: Implement caching, rate limiting, or semantic search
4. Fix Bugs: Report issues or submit pull requests

📝 License

This project is licensed under the MIT License - see the LICENSE file for details.

🙏 Acknowledgments

• FastMCP - MCP server framework
• Serper - Web search API
• Trafilatura - Web content extraction
• Model Context Protocol - MCP specification

📧 Support

Having issues? Here's how to get help:

1. 📖 Check this README thoroughly
2. 🔍 Use MCP Inspector for debugging
3. 💬 Open an issue on GitHub