MCP Client

A Python client for interacting with Claude AI models using the Model Context Protocol (MCP).

Description

This project provides a command-line interface for connecting to MCP servers and using their tools with Claude AI models. It allows you to:

• Connect to MCP servers via Docker, Python, JavaScript, or TypeScript
• Send queries to Claude AI models
• Use tools provided by connected MCP servers
• Process responses in an interactive chat loop

Requirements

• Python 3.7+
• Anthropic API key
• MCP server(s)

Installation

1. Clone this repository
2. Install dependencies:

pip install -r requirements.txt

3. Create a .env file based on .env.example and add your Anthropic API key:

ANTHROPIC_API_KEY=your_anthropic_api_key_here

Configuration

Configure MCP servers by setting environment variables in your .env file:

MCP_MEMORY_SERVER=docker://mcp/memory:latest
# Add more MCP servers as needed
# MCP_WEATHER_SERVER=path_to_weather_server_script

Usage

Run the client:

python mcp_client.py

Enter your questions at the prompt. Type quit or exit to end the session.

Supported MCP Server Types

• Docker images: docker://<image[:tag]>
• Python scripts: *.py
• JavaScript files: *.js
• TypeScript files: *.ts

License

[License information]