Weather MCP Server (@philschmid/weather-mcp)

A simple Model Context Protocol (MCP) server that provides weather forecast information.

Description

This server implements the Model Context Protocol to expose a single tool: get_weather_forecast.

When called, this tool:

1. Takes a location (e.g., "London, UK") and a date (e.g., "2024-07-26") as input.
2. Uses node-geocoder with the OpenStreetMap provider to find the latitude and longitude for the given location.
3. Fetches the hourly temperature forecast for the specified date from the Open-Meteo API.
4. Returns the forecast as a JSON object mapping timestamps to temperatures (in Celsius).

Installation & Running

This server is designed to be run directly using npx. Ensure you have Node.js and npm installed.

npx -y @philschmid/weather-mcp

This command will download (if necessary) and run the server, making it available for MCP clients to connect to via standard input/output (stdio).

Usage with an MCP Client (Python Example)

You can interact with this server using any MCP-compatible client. Here's an example using Python with the mcp library and Google's Generative AI SDK to utilize the server's tool via function calling:

import asyncio
import os
from datetime import datetime
from google import genai
from google.genai import types
from mcp import ClientSession, StdioServerParameters
from mcp.client.stdio import stdio_client

# Configure Gemini client (replace with your API key setup)
client = genai.Client(
    api_key=os.getenv("GEMINI_API_KEY")
)

# Server parameters for stdio connection
server_params = StdioServerParameters(
    command="npx",  # Executable
    args=[
        "-y",
        "@philschmid/weather-mcp", # The published npm package
    ],
    env=None, # Optional environment variables
)

async def run():
    # Connect to the server via stdio
    async with stdio_client(server_params) as (read, write):
        # Start an MCP client session
        async with ClientSession(
            read,
            write,
        ) as session:
            prompt = f"What is the weather in London, UK for {datetime.now().strftime('%Y-%m-%d')}?"

            # Initialize the connection
            await session.initialize()
            print(f"MCP Session Initialized with server: {session.server_info.name} v{session.server_info.version}")

            # Get tools from the server
            mcp_tools = await session.list_tools()

            # Convert MCP tools to Gemini Tool format
            gemini_tools = [
                types.Tool(
                    function_declarations=[
                        types.FunctionDeclaration(
                            name=tool.name,
                            description=tool.description,
                            parameters=types.Schema(
                                type=types.Type.OBJECT,
                                properties={
                                    k: types.Schema(
                                         type=types.Type.STRING, # Simplified for example
                                         description=v.get('description', '')
                                     )
                                    for k, v in tool.inputSchema.get('properties', {}).items()
                                },
                                required=tool.inputSchema.get('required', [])
                            )
                        )
                    ]
                )
                for tool in mcp_tools.tools
            ]
            print(f"Tools registered: {[t.function_declarations[0].name for t in gemini_tools]}")


            # Send prompt to Gemini, making it aware of the server's tools
            print(f"Sending prompt to Gemini: '{prompt}'")
            response = client.models.generate_content(
                model="gemini-pro", # Or another suitable Gemini model
                contents=[prompt],
                generation_config=types.GenerateContentConfig(
                    temperature=0,
                ),
                 tools=gemini_tools,
            )

            # Check if Gemini decided to call the weather tool
            if response.candidates[0].content.parts[0].function_call:
                function_call = response.candidates[0].content.parts[0].function_call
                print(f"Gemini Function Call: {function_call.name}({function_call.args})")

                # Execute the function call using the MCP server
                result = await session.call_tool(
                    function_call.name, arguments=dict(function_call.args)
                )
                weather_data = result.content[0].text
                print(f"Weather Server Response: {weather_data}")

                # (Optional) Send the result back to Gemini to get a natural language response
                # ...

            else:
                # Gemini responded directly without using the tool
                print("Gemini response (no function call):")
                print(response.text)


# Run the asynchronous function
if __name__ == "__main__":
    # Make sure to set your GEMINI_API_KEY environment variable
    if not os.getenv("GEMINI_API_KEY"):
        print("Error: GEMINI_API_KEY environment variable not set.")
    else:
        try:
            asyncio.run(run())
        except Exception as e:
            print(f"An error occurred: {e}")

This project is licensed under the MIT License - see the LICENSE file for details.