MCP Demo

What is MCP Demo?

MCP Demo is a demonstration MCP (Model Context Protocol) server that provides weather forecasts and alerts from the U.S. National Weather Service. It is designed for use with language models like DeepSeek-V3 (deepseek-chat) that support the tools attribute.

How to use MCP Demo?

Clone the repository, create a Python virtual environment with uv, and set your OPENAI_API_KEY in a .env file. Run the client with python client/main.py server/main.py. The client accepts natural language queries about weather.

Key features of MCP Demo

• Provides two tools: get_alerts and get_forecast
• Uses the free National Weather Service API (api.weather.gov)
• Requires DeepSeek-V3 model (deepseek-chat) for tool support
• Written in Python, with a Golang version also available
• Simple client-server architecture over stdio

Use cases of MCP Demo

• Get current weather forecast for any U.S. city
• Retrieve weather alerts for a given location
• Demonstrate MCP integration with an LLM

FAQ from MCP Demo

What tools are available?

Two tools: get_alerts and get_forecast.

How do I install and run it?

Install uv, clone the repo, create a virtual environment, and run python client/main.py server/main.py. Set your OPENAI_API_KEY in a .env file first.

Which LLM does this demo work with?

It works with deepseek-chat (DeepSeek-V3) because only V3 has the tools attribute. Other models like deepseek-reasoner are not supported.

Do I need any API keys?

Yes, you need an OpenAI API key (though the demo queries the NWS API, not OpenAI). Set it in the .env file.

Where does the weather data come from?

The data comes from the public U.S. National Weather Service API at https://api.weather.gov.