🌤️ MCP Weather Server

A production-ready Model Context Protocol weather server built with the official Anthropic FastMCP Python SDK. Provides weather forecasts, alerts, and current conditions through a standardized AI-agent interface.

✨ Features

• 🔧 2 Tools: get_alerts (weather alerts by state) and get_forecast (5-day forecast by coordinates)

• 📄 1 Resource: weather://{state}/current — formatted current conditions

• 💬 1 Prompt: weather_assistant(state) — guided weather analysis template

• 🔌 Dual Transport: stdio (local clients) and Streamable HTTP (remote clients)

• 🪵 Structured Logging: All server logs go to stderr (compatible with stdio transport)

• 📦 No API Keys Required: Uses realistic simulated weather data

• 🧪 Test Client Included: Ready-to-run example client

Related MCP server: MCP Weather Server

🚀 Quick Start

Option 1: pip

python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
python weather_server.py

Option 2: uv (recommended)

uv venv
source .venv/bin/activate
uv add "mcp[cli]>=1.27,<2" httpx
python weather_server.py

Option 3: Setup script

chmod +x setup.sh
./setup.sh
python weather_server.py

📋 Usage

stdio Transport (default)

python weather_server.py

The server listens on stdin/stdout, ready to connect to Claude Desktop, Cursor, or any MCP client.

Streamable HTTP Transport

python weather_server.py --transport streamable-http --port 8000

Now available at http://localhost:8000/mcp for HTTP-based clients.

Run the Test Client

# In one terminal:
python weather_server.py

# In another terminal:
source .venv/bin/activate
python client_example.py

🛠️ Tools Reference

get_alerts(state: str) -> list[dict]

Get active weather alerts for a US state.

Returns: List of alerts with type, severity, and area.

get_forecast(latitude: float, longitude: float) -> list[dict]

Get a 5-day weather forecast for coordinates.

Returns: 5-day forecast with date, temperature_f, condition, humidity_pct, wind_mph.

📄 Resources Reference

weather://{state}/current

Returns formatted current weather conditions for a state as markdown.

💬 Prompts Reference

weather_assistant(state: str)

Creates a weather analysis assistant prompt template.

🔌 Connecting to Clients

Claude Desktop

Add to your claude_desktop_config.json:

{
  "mcpServers": {
    "weather": {
      "command": "python",
      "args": ["/ABSOLUTE/PATH/TO/mcp-weather-server/weather_server.py"]
    }
  }
}

Cursor

In Cursor Settings → Features → MCP Servers, add:

• Name: weather

• Type: command

• Command: python /ABSOLUTE/PATH/TO/mcp-weather-server/weather_server.py

HTTP Client (Streamable HTTP)

curl -X POST http://localhost:8000/mcp \
  -H "Content-Type: application/json" \
  -d '{"jsonrpc":"2.0","id":"1","method":"tools/list","params":{}}'

📁 Project Structure

mcp-weather-server/
├── weather_server.py      # Main server implementation
├── client_example.py      # Example MCP client
├── requirements.txt       # Python dependencies
├── setup.sh              # Automated setup script
├── README.md             # This file
└── .gitignore

🧠 Architecture

┌─────────────┐     stdio/HTTP      ┌──────────────────┐
│  MCP Client  │ ◄──────────────►  │  Weather Server   │
│ (Claude, etc)│                    │  (FastMCP)        │
└─────────────┘                     └──────────────────┘
                                           │
                                    ┌──────┴──────┐
                                    │  Simulated   │
                                    │  Weather Data│
                                    └─────────────┘

📜 License

MIT — see LICENSE for details.

🙌 Contributing

PRs welcome! Please ensure tools have proper type hints and docstrings.