MCP Client

What is MCP Client?

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

How to use MCP Client?

Install dependencies with uv pip install -r requirements.txt or uv pip install, set up an ANTHROPIC_API_KEY in a .env file, then run uv run client.py ./server/weather.py (or python client.py ./server/weather.py) to start the interactive chat. Type queries or quit to exit.

Key features of MCP Client

• Interactive chat client for MCP servers
• Example weather server with US state alerts and forecasts
• Extensible: connect to any MCP-compatible server script
• Uses Anthropic Claude API for natural language understanding
• Supports .env file for API key configuration

Use cases of MCP Client

• Query weather alerts for a US state via natural language
• Get location-based weather forecasts using NWS API
• Test or interact with any custom MCP server script
• Prototype AI‑powered tools that leverage MCP infrastructure

FAQ from MCP Client

What are the runtime requirements?

Python 3.12+, mcp>=1.9.2, and a valid Anthropic API key are required. uv is recommended but optional.

How does MCP Client connect to MCP servers?

It launches the server script as a subprocess (e.g., ./server/weather.py) and communicates via the MCP protocol. The client discovers server tools and processes natural language queries through Claude.

Where is the API key stored?

In a .env file at the project root (ANTHROPIC_API_KEY=your-key). It is loaded automatically and excluded from version control via .gitignore.

Does the weather server require internet?

Yes, it needs to reach https://api.weather.gov for alerts and forecasts.

What transport does the client use?

The client connects to MCP servers over stdio (standard input/output) by running the server script locally.