RatnadeepSimhadri/weather-mcp-server

weather-mcp-server

MCP Weather Server Explanation

This file implements a Model Context Protocol (MCP) server that provides weather information tools. MCP enables AI models to interact with external tools in a structured way.

Overview The server exposes two weather tools:

• get-alerts: Gets weather alerts for a specific US state
• get-forecast: Gets weather forecast for a specific location by coordinates It uses the National Weather Service (NWS) API to fetch real weather data.

How MCP Clients Understand Tool Inputs

The key to how MCP clients understand tool inputs lies in how tools are registered:

server.tool(
  "get-forecast",                   // Tool name
  "Get weather forecast for a location",  // Tool description
  {                                // Input schema
    latitude: z.number().min(-90).max(90).describe("Latitude of the location"),
    longitude: z.number().min(-180).max(180).describe("Longitude of the location"),
  },
  async ({ latitude, longitude }) => {   // Handler function
    // Implementation
  }
);

Here's how clients understand the inputs:

1. Schema Definition: The third parameter uses Zod (z) to define the expected inputs and their constraints:

• Input parameter names (e.g., latitude, longitude)
• Data types (z.number(), z.string())
• Validation rules (min(-90), max(90), length(2))
• Descriptions (describe("Latitude of the location"))

2. Client Discovery: When an MCP client connects to this server:

• It receives metadata about available tools
• For each tool, it gets the name, description, and input schema
• The client can use this information to validate inputs before sending them

3. Communication: The StdioServerTransport handles the serialization and communication of this schema information between the server and client.

This schema-based approach ensures that clients know exactly what parameters each tool expects, their types, constraints, and what they represent.