Microsoft published LangChain.js for Beginners, a free, open-source course with 8 chapters and 70+ runnable TypeScript examples. Clone the repo, drop your API key in a .env file, and you’re building. No Python required.
The course skips the usual tutorial order. Instead of starting with document loaders and embeddings, it gets to tools and agents by Chapter 4. Chapter 5 covers the ReAct pattern for multi-step reasoning. Chapter 6 covers MCP — the Model Context Protocol — including how to build MCP servers and connect agents over both HTTP and stdio transports. RAG doesn’t show up until Chapters 7 and 8, by which point you already understand why an agent would choose to search versus answer from memory. That sequencing matters: agentic RAG with selective retrieval is faster and cheaper than querying embeddings on every turn.
TIP
The course is provider-agnostic. Set four environment variables (AI_API_KEY, AI_ENDPOINT, AI_MODEL, AI_EMBEDDING_MODEL) and every example runs against GitHub Models (free), Microsoft Foundry, or OpenAI — no code changes needed.
The capstone project is an MCP-powered RAG server that exposes document search and ingestion as MCP tools over HTTP, letting multiple agents share a single knowledge base. Bonus samples in the README go further: a burger-ordering agent with a serverless API, a serverless AI chat with RAG on Azure, and a multi-agent travel planner across Azure Container Apps.
If you know npm install and async/await, this is a direct path to production-style agent architecture. Start at github.com/microsoft/langchainjs-for-beginners. If generative AI fundamentals are still fuzzy, Microsoft also has a companion course — Generative AI with JavaScript — worth hitting first.