A hands-on course for building MCP (Model Context Protocol) servers, guided by AI. Learn by doing — Claude teaches the concepts, you write the code.
MCP is an open protocol by Anthropic that lets AI assistants connect to external tools and data sources. Think of it as building APIs that AI models can use — instead of REST endpoints for browsers, you create tools, resources, and prompts for AI clients like Claude.
- Experience: Senior-level JavaScript/TypeScript knowledge
- Node.js: v18+ installed
- Editor: VS Code with the Claude Code extension
- Claude Code: Configured and working in VS Code
- Clone this repository
- Open it in VS Code
- Start a Claude Code chat
- Say: "Start lesson 1"
Claude will read the lesson plan and guide you step-by-step through building each server. No copy-pasting walls of code — you'll learn the concepts and write the implementation yourself.
| Lesson | Server | What You'll Learn |
|---|---|---|
| 01 | Hello MCP | Server basics, tools, Zod schemas, stdio transport, MCP Inspector |
| 02 | Markdown Notes | Resources, reading files, URI schemes |
| 03 | Weather Lookup | External API calls, input validation, error handling |
| 04 | Prompt Library | Prompt templates, dynamic arguments |
| 05 | WordPress Toolkit | Combining tools + resources, real-world CMS integration |
| 06 | Project Scaffolder | Multi-step tools, file system operations, complex inputs |
Each lesson builds on the previous one. Start from lesson 1 if you're new to MCP.
lessons/ # Lesson outlines that guide the AI instructor
servers/ # Your code — each lesson creates a server here
CLAUDE.md # AI instructor configuration
You can jump to any lesson by telling Claude:
- "Start lesson 3"
- "I'd like to work on the weather lookup server"
- "Continue where I left off"
Claude will check your servers/ directory to see what you've already built and pick up from there.
- TypeScript (ES modules)
@modelcontextprotocol/sdk— Official MCP SDKzod— Input validation- Node.js runtime