MCP Documentation Server

A powerful Model Context Protocol (MCP) server that enables semantic search across technical documentation. Built with FastMCP and advanced natural language processing, this server allows you to upload .txt and .md files and perform intelligent semantic searches to quickly find relevant information.

✨ Features

🧠 Semantic Search Engine

• Advanced NLP: Uses paraphrase-multilingual-mpnet-base-v2 model for high-quality embeddings
• Intelligent Chunking: Smart text segmentation with overlap for context preservation
• Multilingual Support: Works with Italian, English, and other languages
• Lightning Fast: Sub-60ms search times after initial processing

📁 Document Management

• Multiple Formats: Support for .txt and .md files
• Automatic Processing: Scans directory for new/modified documents
• Metadata Tracking: File size, modification dates, character counts
• Cache System: Efficient embedding storage and retrieval

🔧 MCP Integration

• Standard Compliant: Full MCP protocol support
• Four Core Tools: Document upload, listing, searching, and removal
• Async Architecture: Non-blocking operations for better performance
• Error Handling: Robust error management and logging

📊 Proven Performance

• Tested at Scale: Successfully handles 639KB+ documents (19,000+ lines)
• 100% Success Rate: All test queries return relevant results
• High Accuracy: Average similarity scores > 0.66
• Production Ready: Thoroughly tested with real technical documentation

🚀 Quick Start

Automated Installation (Recommended)

The easiest way to install and configure the server is using our automated installer:

# Install the package
pip install mcp-documentation-server

# Run the automated setup
mcp-doc-install

What it does:

• ✅ Creates an isolated virtual environment (no permission issues!)
• ✅ Installs the server and all dependencies
• ✅ Sets up directories and configuration
• ✅ Generates wrapper scripts for easy CLI access
• ✅ Creates configuration files for all major MCP clients
• ✅ Provides step-by-step setup instructions

Manual Installation Options

Option 1: Install from PyPI

pip install mcp-documentation-server

Option 2: Install from Source

git clone <repository-url>
cd mcp-documentation-server
pip install -e .

Setup and Configuration

Automatic Setup (Recommended)

Run the installation wizard to configure the server for your MCP client:

mcp-doc-install

This will:

• Create an isolated virtual environment in ~/.mcp-documentation-server/venv
• Install the server without requiring system-wide permissions
• Create necessary directories
• Generate configuration files for Claude Desktop, Cursor, and VS Code
• Create wrapper scripts for easy command-line usage
• Provide step-by-step setup instructions

Manual Setup

1. Create a base directory for your documentation:

   mkdir ~/.mcp-documentation-server
   cd ~/.mcp-documentation-server
   mkdir data embeddings
   

2. Configure your MCP client (see Configuration section below)

3. Add your documents to the data directory

Quick Test

1. After automated setup, use the wrapper scripts:

   # Windows
   C:\Users\YourName\.mcp-documentation-server\scripts\mcp-documentation-server.bat
   
   # Linux/Mac
   ~/.mcp-documentation-server/scripts/mcp-documentation-server
   

2. Or start the server directly (if installed globally):

   mcp-documentation-server
   

3. Or use the launcher with custom settings:

   mcp-doc-launcher --base-dir ~/my-docs --log-level DEBUG
   

⚙️ Configuration

Environment Variables

The server can be configured using environment variables:

• MCP_DOC_BASE_DIR: Base directory for the server (default: current directory)
• MCP_DOC_DATA_DIR: Directory containing documentation files (default: data)
• MCP_DOC_EMBEDDINGS_DIR: Directory for embeddings cache (default: embeddings)
• MCP_DOC_METADATA_FILE: Metadata file path (default: metadata.json)
• MCP_DOC_LOG_LEVEL: Logging level (DEBUG/INFO/WARNING/ERROR, default: INFO)

MCP Client Configuration

Claude Desktop

Add to your Claude Desktop configuration file:

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

{
  "mcpServers": {
    "documentation": {
      "command": "mcp-documentation-server",
      "env": {
        "MCP_DOC_BASE_DIR": "/path/to/your/documentation",
        "MCP_DOC_LOG_LEVEL": "INFO"
      }
    }
  }
}

Cursor

Add to your Cursor MCP configuration:

{
  "mcpServers": {
    "documentation": {
      "command": "mcp-documentation-server",
      "env": {
        "MCP_DOC_BASE_DIR": "/path/to/your/documentation",
        "MCP_DOC_LOG_LEVEL": "INFO"
      }
    }
  }
}

VS Code with MCP Extension

Add to your VS Code MCP configuration:

{
  "mcpServers": {
    "documentation": {
      "command": "mcp-documentation-server", 
      "env": {
        "MCP_DOC_BASE_DIR": "/path/to/your/documentation",
        "MCP_DOC_LOG_LEVEL": "INFO"
      }
    }
  }
}

For detailed VS Code integration instructions, see VS_CODE_INTEGRATION.md.

🎯 Why Automated Installation?

Our automated installer (mcp-doc-install) provides significant advantages:

🔒 Isolated Environment

• No System Pollution: Installs in a dedicated virtual environment
• No Permission Issues: Works without administrator/sudo privileges
• Clean Uninstall: Easy to remove without affecting your system

🚀 Zero Configuration

• Smart Defaults: Automatically detects optimal settings
• Cross-Platform: Works on Windows, macOS, and Linux
• Multiple Clients: Generates configs for Claude, Cursor, and VS Code simultaneously

🛠️ Developer Friendly

• Wrapper Scripts: Easy-to-use command-line tools
• Local Development: Automatically finds and installs local builds
• Professional Setup: Follows Python packaging best practices

📁 Organized Structure

~/.mcp-documentation-server/
├── venv/                    # Isolated Python environment
├── data/                    # Your documentation files
├── embeddings/              # Cached embeddings for fast search
├── configs/                 # Generated MCP client configs
└── scripts/                 # Wrapper scripts for easy CLI access

📋 Usage

MCP Tools

Once configured with your MCP client, you can use these tools:

1. get_upload_info: Get information about where to upload documents
2. list_documents: List all uploaded documents with metadata
3. search_documents: Perform semantic search across all documents
4. remove_document: Remove a document from the index

Example Queries

With your MCP client (Claude Desktop, etc.), you can ask:

• "What documents do I have available?"
• "Search for information about authentication"
• "Find documentation about API endpoints"
• "How do I configure the database connection?"

Command Line Usage

Interactive Mode

# Start with default settings
mcp-documentation-server

# Start with custom configuration
mcp-doc-launcher --base-dir ~/my-docs --log-level DEBUG

Environment Variables

export MCP_DOC_BASE_DIR=~/my-documentation
export MCP_DOC_LOG_LEVEL=DEBUG
mcp-documentation-server

.\venv\Scripts\activate

macOS/Linux

source venv/bin/activate

3. **Install dependencies**
```bash
pip install -r requirements.txt

4. Start the server

   python server.py
   

Adding Documents

Simply place your .txt or .md files in the data/ directory. The server will automatically detect and process them on the next operation.

🛠️ MCP Tools

The server provides four main tools for document management and search:

1. get_upload_info

Returns information about where and how to upload documents.

Example Response:

{
  "upload_path": "/path/to/data",
  "supported_formats": [".txt", ".md"],
  "instructions": "Place files in the data directory..."
}

2. list_documents

Lists all available documents with metadata.

Example Response:

{
  "documents": [
    {
      "id": "technical_docs.md",
      "name": "technical_docs.md",
      "size": 639757,
      "char_count": 619348,
      "last_modified": "2025-06-14T00:30:00"
    }
  ]
}

3. search_documents

Performs semantic search across all documents.

Parameters:

• query (string): Search query
• max_results (integer, optional): Maximum results to return (default: 5)

Example:

{
  "query": "Docker containerization",
  "max_results": 3
}

Response:

{
  "results": [
    {
      "document_id": "technical_docs.md",
      "similarity_score": 0.8423,
      "text": "Docker is a containerization platform...",
      "chunk_id": 5,
      "start_char": 1250,
      "end_char": 1750
    }
  ],
  "query": "Docker containerization",
  "total_results": 15
}

4. remove_document

Removes a document and its associated embeddings.

Parameters:

• filename (string): Name of the file to remove

📈 Performance Benchmarks

Based on comprehensive testing with real technical documentation:

Metric	Value	Notes
Document Size Tested	639 KB	MCP protocol documentation
Search Success Rate	100%	All 21 test queries successful
Average Search Time	50.7ms	After initial processing
Average Similarity Score	0.6689	High relevance results
Best Match Score	0.8423	"MCP Model Context Protocol"
Processing Time	~3min	639KB initial processing

Test Categories Validated

• ✅ Basic Concepts
• ✅ Client Implementations
• ✅ Tool Development
• ✅ Resource Management
• ✅ Prompt Engineering
• ✅ Technical Specifications

⚙️ Configuration

Environment Settings

The server uses the following directory structure:

mcp-documentation-server/
├── data/              # Place your documents here
├── embeddings/        # Cached embeddings (auto-generated)
├── venv/             # Virtual environment
├── server.py         # Main MCP server
├── document_manager.py # Document handling
├── search_engine.py  # Semantic search engine
└── metadata.json     # Document metadata

Customization Options

Model Selection: Change the embedding model in search_engine.py:

model_name = "paraphrase-multilingual-mpnet-base-v2"  # Default
# model_name = "all-MiniLM-L6-v2"  # Faster, smaller

Chunk Settings: Adjust chunking parameters:

chunk_size = 500      # Characters per chunk
overlap = 50         # Overlap between chunks

Search Results: Modify default result count:

default_max_results = 5  # Default results returned

� Integration Examples

Claude Desktop Configuration

Add to your Claude Desktop MCP settings:

{
  "mcpServers": {
    "documentation-search": {
      "command": "python",
      "args": ["/path/to/mcp-documentation-server/server.py"],
      "cwd": "/path/to/mcp-documentation-server"
    }
  }
}

VS Code + GitHub Copilot Integration

Prerequisites: VS Code 1.99+, GitHub Copilot subscription

1. Enable MCP Support in VS Code settings:

   {
     "chat.mcp.enabled": true,
     "chat.mcp.discovery.enabled": true
   }
   

2. Configure Server - The .vscode/mcp.json file is already included:

   {
     "servers": {
       "documentationSearch": {
         "type": "stdio",
         "command": "python",
         "args": ["server.py"],
         "cwd": "${workspaceFolder}"
       }
     }
   }
   

3. Use in Copilot Chat:

   • Open Chat (Ctrl+Alt+I)
   • Switch to Agent Mode
   • Select Tools and enable documentation search tools
   • Ask: Search for Docker containerization in my docs

📖 Full Setup Guide: See VS_CODE_INTEGRATION.md

Using with Other MCP Clients

The server follows standard MCP protocol and works with any compliant client:

1. Start the server via stdio transport
2. Connect your client to the server process
3. Use the tools for document search and management

🧪 Testing

Run the included test suite to verify functionality:

# Basic functionality test
python test_direct.py

# Advanced testing with large documents
python test_advanced.py

# Quick search verification
python test_quick.py

📋 Requirements

Python Dependencies

fastmcp>=0.1.0
sentence-transformers>=2.2.2
scikit-learn>=1.3.0
numpy>=1.24.0

System Requirements

• RAM: 2GB+ (model loading)
• Storage: 500MB+ (model cache)
• CPU: Any modern processor

🤝 Contributing

1. Fork the repository
2. Create a feature branch: git checkout -b feature/amazing-feature
3. Make your changes and add tests
4. Commit your changes: git commit -m 'Add amazing feature'
5. Push to the branch: git push origin feature/amazing-feature
6. Open a Pull Request

Development Setup

# Install development dependencies
pip install -r requirements-dev.txt

# Run tests
python -m pytest tests/

# Code formatting
black .
flake8 .

📄 License

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

🙋‍♂️ Support

• Issues: Report bugs and request features on GitHub Issues
• Documentation: Check the docs/ directory for detailed guides
• Community: Join discussions in GitHub Discussions

🔮 Roadmap

Planned Features

• PDF Support: Add PDF document processing
• Web Interface: Optional web UI for document management
• Docker Support: Containerized deployment
• Advanced Filters: Search by document type, date, size
• Hybrid Search: Combine semantic and keyword search
• REST API: Additional HTTP API alongside MCP

Performance Improvements

• Incremental Updates: Update only changed document sections
• Compressed Embeddings: Reduce storage requirements
• Distributed Processing: Multi-threaded document processing

---

📊 Project Stats

• Languages: Python
• Framework: FastMCP
• AI Model: Sentence Transformers
• Protocol: Model Context Protocol (MCP)
• Status: ✅ Production Ready
• Test Coverage: 100% core functionality
• Documentation: Complete

Built with ❤️ for the MCP community