MCP Client

A Python client for interacting with MCP servers, featuring a weather tool server as an example. The client uses Anthropic's Claude API for natural language queries and can connect to any compatible MCP server script.

Features

• Interactive chat client for MCP servers
• Example weather server (server/weather.py) with:
  • US state weather alerts
  • Location-based weather forecasts (via NWS API)
• Extensible: connect to any MCP-compatible server script
• Uses Anthropic Claude API for natural language understanding

Requirements

• Python 3.12+
• Anthropic API key
• MCP Python package (mcp>=1.9.2)
• uv (for running with isolated environments, optional but recommended)

Installation

1. Clone the repository:

   git clone <your-repo-url>
   cd mcp-client
   

2. Install dependencies:

   uv pip install -r requirements.txt
   

   Or, if you use uv and pyproject.toml:

   uv pip install
   

3. Set up your Anthropic API key:

   • Create a .env file in the project root:

     ANTHROPIC_API_KEY=your-anthropic-api-key-here
     

   • The .env file is in .gitignore and will be loaded automatically.

Usage

Start the Weather Server

The example weather server is located at server/weather.py. It provides two tools: get_alerts and get_forecast.

Run the Client

You can connect the client to the weather server using:

uv run client.py ./server/weather.py

Or, if you prefer plain Python:

python client.py ./server/weather.py

You should see output like:

Connected to server with tools: ['get_alerts', 'get_forecast']

MCP Client Started!
Type your queries or 'quit' to exit.

Example Queries

• What is the weather in New York?
• Are there any weather alerts for CA?
• Give me the forecast for latitude 40.7, longitude -74.0

Exiting

Type quit to exit the chat loop.

Project Structure

.
├── client.py            # Main MCP client
├── server/
│   └── weather.py       # Example weather MCP server
├── pyproject.toml       # Project metadata and dependencies
├── .env                 # (Not committed) Your API keys
├── .gitignore
└── README.md

Development

• The client loads environment variables from .env using python-dotenv.
• The weather server uses the US National Weather Service API and requires internet access.

Troubleshooting

• Anthropic API errors:
  Ensure your API key is valid and has sufficient credits.
• Python version:
  The project requires Python 3.12 or newer.
• Network issues:
  The weather server needs to reach https://api.weather.gov.

License