FastAPI MCP Server with Weather Tools

A Model Context Protocol (MCP) server built with FastAPI that provides weather information using the National Weather Service API. Features streamable HTTP transport for real-time communication with MCP Inspector.

🌐 Azure Deployment

• Azure URL: https://<APP-SERVICE-NAME>.azurewebsites.net/
• API Documentation: https://<APP-SERVICE-NAME>.azurewebsites.net/docs
• Health Check: https://<APP-SERVICE-NAME>.azurewebsites.net/
• MCP Endpoint: https://<APP-SERVICE-NAME>.azurewebsites.net/mcp/stream

You can test the weather tools without local setup:

# Test weather alerts for California
curl -X POST "https://`<APP-SERVICE-NAME>`.azurewebsites.net/mcp/stream" \
  -H "Content-Type: application/json" \
  -d '{"jsonrpc": "2.0", "id": 1, "method": "tools/call", "params": {"name": "get_alerts", "arguments": {"state": "CA"}}}'

# Test weather forecast for San Francisco
curl -X POST "https://`<APP-SERVICE-NAME>`.azurewebsites.net/mcp/stream" \
  -H "Content-Type: application/json" \
  -d '{"jsonrpc": "2.0", "id": 2, "method": "tools/call", "params": {"name": "get_forecast", "arguments": {"latitude": 37.7749, "longitude": -122.4194}}}'

Features

• FastAPI Framework: Modern, fast web framework for building APIs
• MCP Protocol Compliance: Full support for JSON-RPC 2.0 MCP protocol
• Streamable HTTP Transport: HTTP-based streaming for MCP Inspector connectivity
• Weather Tools:
  • get_alerts: Get weather alerts for any US state
  • get_forecast: Get 5-day weather forecast for any location (latitude/longitude)
• Sample Resources: Basic resource handling demonstration
• Virtual Environment: Properly isolated Python environment
• Auto-reload: Development server with automatic reloading
• National Weather Service API: Real-time weather data from official US government source

Prerequisites

• Python 3.8+
• pip (Python package installer)

Setup

1. Create and activate virtual environment:

   python -m venv venv
   .\venv\Scripts\Activate.ps1
   

2. Install dependencies:

   pip install -r requirements.txt
   

3. Start the server:

   .\start_server.ps1
   

   Or manually:

   .\venv\Scripts\python.exe -m uvicorn main:app --host 0.0.0.0 --port 8000 --reload
   

Connecting to MCP Inspector

Method 1: Azure-Hosted Server (No Setup Required)

Connect directly to the live Azure deployment:

Configuration for MCP Inspector:

{
  "mcpServers": {
    "weather-mcp-server-azure": {
      "transport": {
        "type": "http",
        "url": "https://`<APP-SERVICE-NAME>`.azurewebsites.net/mcp/stream"
      },
      "name": "Weather MCP Server (Azure)",
      "description": "MCP Server with weather forecast and alerts tools hosted on Azure"
    }
  }
}

Method 2: Local Development Server

1. Start the MCP server (it will run on http://localhost:8000)

2. In MCP Inspector v0.13.0:

   • Add a new server connection
   • Use HTTP transport type
   • URL: http://localhost:8000/mcp/stream

3. Configuration file (mcp-config.json):

   {
     "mcpServers": {
       "weather-mcp-server-local": {
         "transport": {
           "type": "http",
           "url": "http://localhost:8000/mcp/stream"
         },
         "name": "Weather MCP Server (Local)",
         "description": "MCP Server with weather forecast and alerts tools running locally"
       }
     }
   }
   

Method 3: Web Test Interface

Visit http://localhost:8000/test (local) or https://<APP-SERVICE-NAME>.azurewebsites.net/test (Azure) to use the built-in web interface for testing HTTP connectivity.

API Endpoints

• GET /health: Server health check
• POST /mcp/stream: Main MCP endpoint with streamable HTTP
• GET /mcp/capabilities: Get server capabilities
• GET /test: Web-based HTTP test interface
• POST /mcp: HTTP MCP endpoint (legacy)

Usage

Start the server:

pip install -r requirements.txt
python main.py

The server will start on http://localhost:8000

Example MCP requests:

Initialize

{
  "jsonrpc": "2.0",
  "method": "initialize",
  "params": {},
  "id": 1
}

List Tools

{
  "jsonrpc": "2.0",
  "method": "tools/list",
  "params": {},
  "id": 2
}

Call Weather Alert Tool

{
  "jsonrpc": "2.0",
  "method": "tools/call",
  "params": {
    "name": "get_alerts",
    "arguments": {
      "state": "CA"
    }
  },
  "id": 3
}

Call Weather Forecast Tool

{
  "jsonrpc": "2.0",
  "method": "tools/call",
  "params": {
    "name": "get_forecast",
    "arguments": {
      "latitude": 37.7749,
      "longitude": -122.4194
    }
  },
  "id": 4
}

Testing

Test with Python client:

.\venv\Scripts\python.exe test_http_client.py  # Streamable HTTP client

Test with web interface:

Open http://localhost:8000/test in your browser

Available Tools

1. get_alerts: Get weather alerts for a US state

   {
     "name": "get_alerts",
     "arguments": {
       "state": "CA"
     }
   }
   

   • Parameter: state (string) - Two-letter US state code (e.g., CA, NY, TX)
   • Returns: Active weather alerts including severity, description, and instructions

2. get_forecast: Get weather forecast for a location

   {
     "name": "get_forecast", 
     "arguments": {
       "latitude": 37.7749,
       "longitude": -122.4194
     }
   }
   

   • Parameters:
     • latitude (number) - Latitude coordinate
     • longitude (number) - Longitude coordinate
   • Returns: 5-day weather forecast with temperature, wind, and detailed conditions

Weather Data Source

This server uses the National Weather Service (NWS) API, which provides:

• Real-time weather alerts and warnings
• Detailed weather forecasts
• Official US government weather data
• No API key required
• High reliability and accuracy

Available Resources

• mcp://server/sample: Sample resource for demonstration

Troubleshooting

MCP Inspector Connection Issues:

1. Ensure the server is running on http://localhost:8000
2. Verify MCP endpoint is accessible: http://localhost:8000/mcp/stream
3. Check capabilities endpoint: http://localhost:8000/mcp/capabilities
4. Try the web test interface first: http://localhost:8000/test

Common Issues:

• Port already in use: Change the port in startup commands
• Virtual environment not activated: Run .\venv\Scripts\Activate.ps1
• Dependencies missing: Run pip install -r requirements.txt