42. Model Context Protocol (MCP)
Mini-Project: Simulated MCP Server: Exposing a Weather Tool
A FastMCP server that exposes weather lookup and comparison tools via the Model Context Protocol, demonstrating how any MCP-compatible client can discover and call these tools without custom integration code.
What It Is
The Model Context Protocol (MCP) is an open standard created by Anthropic that provides a universal way for AI models to connect to external data sources and tools. MCP defines a client-server architecture where an MCP client (the AI application) communicates with MCP servers (tool/data providers) using a standardized JSON-RPC protocol. This allows any MCP-compatible AI model to use any MCP-compatible tool without custom integration code.
MCP solves the "N x M" integration problem: instead of building custom integrations between N models and M tools, each model implements one MCP client and each tool implements one MCP server.
Who Created It / Governed By
Created by Anthropic (November 2024). Open-source specification hosted on GitHub. Rapidly adopted by the ecosystem -- supported by Claude, Cursor, Windsurf, and many tool providers.
How It Works
sequenceDiagram
participant User
participant Host as MCP Host (Claude, IDE)
participant Client as MCP Client
participant Server as MCP Server (Tool Provider)
User->>Host: Request
Host->>Client: Forward request
Client->>Server: Initialize connection (JSON-RPC)
Server-->>Client: Capability advertisement (tools, resources)
Client->>Server: Tool call
Server-->>Client: Tool result
Client-->>Host: Result
Host-->>User: Response with dataKey Features
| Feature | Description |
|---|---|
| Tools | Server exposes callable functions with JSON schemas |
| Resources | Server exposes read-only data (files, DB records) |
| Prompts | Server provides pre-built prompt templates |
| Transport | Supports stdio (local) and HTTP+SSE (remote) |
| Discovery | Client discovers server capabilities dynamically |
| Security | User consent required for tool execution |