Build a working AI chat agent from scratch and watch each layer come together. You start with a plain model request, wrap it in a browser chat interface, move it onto a LangChain agent, give the agent a real tool, and finish by connecting remote tools over the Model Context Protocol (MCP).
You do not need prior experience with LangChain, Chainlit, or agents. Every step is written out, every phase ends with a checkpoint you can run, and the whole workshop uses Groq's free tier, so there is no paid AI account and no credit card.
Your workshop route
| Chapter | Pages | What you produce | Lesson time |
|---|---|---|---|
| 1. Set up your workshop | 3 | A working connection to Groq | 35 minutes |
| 2. Chat with Chainlit | 2 | A streaming chat app with conversation memory | 25 minutes |
| 3. Meet the agent | 1 | An agent created with create_agent | 15 minutes |
| 4. Add tools | 2 | A weather tool the agent can call | 30 minutes |
| 5. Connect MCP tools | 2 | One agent using local and remote tools | 25 minutes |
Allow about 130 minutes for lessons, 14 minutes for chapter checks, and 11 minutes for the final test. Downloads vary with your connection, so the first chapter can take longer the first time. If you are running this at an event, complete the setup chapter before the session starts and split the rest into two sittings if needed.
Each chapter ends with two or three questions. The final test has ten new questions. For multiple-choice questions, select every correct option. Every question must be correct to pass a check or the final test, and an explanation appears after grading so you can revisit the idea.
How you will work
Every lesson creates one small file and then improves it. Type the code yourself rather than copying if you can; the checkpoint after each phase tells you exactly what to run and what success looks like.
Every lesson includes the complete code for that step, so if you get stuck you can compare your file with the lesson, then return to your own work. The course is self-contained and does not depend on any other repository.
Keep your API keys in a local .env file and never commit it. Your Groq API key is a personal credential, so treat it like a password.
What you will learn
- Create an isolated Python environment and install a pinned, reproducible dependency set
- Connect to Groq through an OpenAI-compatible client
- Build a Chainlit chat app with streaming responses and conversation memory
- Convert the app to a LangChain agent with create_agent
- Define a typed tool with the @tool decorator and show tool calls in the UI
- Connect remote MCP tools and combine them with local tools
Before you start
- A free Groq API key, which you can create without a credit card
- Python 3.10 through 3.13, Git, a terminal, and a code editor such as Visual Studio Code
- No previous LangChain, Chainlit, or agent experience required
Course modules
Set up your workshop
Create an isolated Python environment, install pinned dependencies, and prove you can reach Groq.
Chat with Chainlit
Turn the model connection into a browser chat app, then add streaming replies and conversation memory.
Meet the agent
Replace direct model calls with a LangChain agent created by create_agent, and prepare it for tools.
Add tools
Define a typed weather tool with the @tool decorator and show the agent's tool calls in the chat UI.
Connect MCP tools
Fetch tools from a remote MCP server and combine them with your local weather tool in one agent.