DeepMCP

A flexible and powerful Multi-Computer Programming (MCP) framework for building client-server applications with support for natural language interaction and LLM integration.

Features

• Asynchronous HTTP Server: Support for streaming HTTP communication using modern async patterns
• Mathematical Calculator Tools: Built-in arithmetic operations (addition, subtraction, multiplication, division)
• Dual-Mode Client: Can run as both client and server
• LLM Integration: Connect with OpenAI-compatible APIs for natural language processing
• Simple Fallback Implementation: Minimal version that works without external dependencies

Project Structure

deepmcp/
├── src/
│   ├── deepmcp/
│   │   ├── __init__.py
│   │   ├── client.py      # MCP client implementation with LLM integration
│   │   ├── server.py      # MCP server implementation with calculator tools
│   │   └── main.py        # Main entry point
│   └── tests/
├── pyproject.toml        # Project configuration and dependencies
└── minimal_mcp.py        # Minimal implementation without external dependencies

Installation

Prerequisites

• Python 3.12 or higher
• uv package manager (recommended for faster installations)

Steps

1. Clone the repository:

   git clone <repository-url>
   cd deepmcp
   

2. Install dependencies using uv:

   uv pip install -e .
   

   If you encounter issues with dependencies, try installing them directly:

   uv pip install fastmcp>=2.12.3 mcp>=1.14.0 openai>=1.108.0
   

Usage

Starting the Server

You can start the server in several ways:

1. Using the command-line script:

   deepmcp
   

2. Running the server module directly:

   python -m deepmcp.server
   

3. Using the client in server mode:

   python -m deepmcp.client --server
   

Starting the Client

To start the client and connect to the server:

python -m deepmcp.client

By default, the client connects to http://localhost:8000. You can specify a different server URL:

python -m deepmcp.client http://your-server-url:port

Using the Minimal Implementation

If you encounter dependency issues with the main implementation, you can use the minimal version that doesn't rely on external packages:

python minimal_mcp.py

This runs a simplified version of the MCP server and client with basic calculator functionality.

Troubleshooting

Common Issues

1. ModuleNotFoundError: No module named 'deepmcp'

   • Ensure the package is installed in development mode: uv pip install -e .
   • Check that your Python path includes the project directory
   • Verify that the pyproject.toml has the correct package discovery configuration

2. ModuleNotFoundError: No module named 'fastmcp' or other dependencies

   • Try reinstalling the dependencies: uv pip install fastmcp mcp openai
   • If issues persist, use the minimal implementation: python minimal_mcp.py

3. Connection Issues

   • Ensure the server is running before starting the client
   • Check that the client is connecting to the correct host and port
   • Verify that no firewall is blocking the connection

LLM Integration

The client integrates with OpenAI-compatible APIs (via OpenRouter in the current implementation). To use this feature, you'll need to configure your API key:

1. Open client.py
2. Update the OPENROUTER_API_KEY variable with your own API key
3. You can also change the model parameter to use a different LLM model

License

This project is licensed under the MIT License.

Contributing

Contributions are welcome! Please fork the repository and submit a pull request with your changes.

Acknowledgements

• Based on the Multi-Computer Programming (MCP) framework
• Uses FastMCP for asynchronous server implementation
• Integrates with OpenAI-compatible APIs for natural language processing