The best way to make the modules stick is to build something. These six projects take you from a research agent to a RAG chatbot over your own documents, a human-in-the-loop workflow, a multi-agent team, an MCP server, and finally a containerised agent that you deploy.
Each one is a code-along video, so you can pause, build it yourself and adapt it to your own data. Pick the project that matches the module you've just finished and come back for the rest. The videos use different languages and tools, so treat them as examples of the pattern and swap in the model and framework you prefer.
Build a structured research agent from scratch with Pydantic AI and plain Python, combining web search, document downloading and an LLM summariser, with Pydantic models keeping its outputs predictable.
Build a chat-with-your-PDF app using retrieval-augmented generation: text chunking, embeddings, semantic search and a vector database, with a Node.js backend and an LLM API.
Design a stateful agent that pauses for human approval and then resumes where it left off, first with LangChain and then with LangGraph. This one uses the JavaScript versions with Node.js.
Build a small crew of agents that plan, write and edit a blog post, defining each agent's role and task and running them together with CrewAI and a local Ollama model.
Build an MCP server in Python that fetches real-time data and edits Google Sheets, then connect it to Claude Desktop and use it as an AI agent.
The capstone: package an agent with Docker, serve it through FastAPI, deploy it to a hosting platform, and extend it with LangChain and LangGraph tools and a supervisor that coordinates several agents.