What you'll build

You'll build up Atlas, a trip-planning research agent, across the whole course — starting from a single Claude call, adding a tool-calling loop, well-designed tools, memory, planning and reflection, retrieval, and finally multi-agent collaboration. Every module ends with a guided project, and the course closes with a capstone that assembles the full agent with memory and retrieval. The local Git-style mechanics and orchestration are verified for real; the Claude API code uses the low-cost claude-haiku-4-5 model so you can follow along for pennies.

Before you start

You'll need comfortable Python — functions, dictionaries, classes — and a basic familiarity with calling an LLM API. Our Generative AI & LLM Engineering course is the ideal prerequisite: it covers prompting, structured outputs, tool use, and retrieval, which this course builds directly on. If you're new to Python, start with Python for Data Analytics. You'll write real code against the Anthropic API; the examples use the low-cost claude-haiku-4-5 model so you can follow along for pennies.

Set up your environment

You can complete this course on any machine with Python 3.12+. Set up once and you're ready for every module.

1. Get an Anthropic API key from the Anthropic Console, then store it in your environment as ANTHROPIC_API_KEY — never in your code.
2. Install the packages the course uses:

pip install anthropic==0.113.0 pydantic sentence-transformers chromadb

Expected API spend for the whole course is well under $1 on the default claude-haiku-4-5 model. The retrieval module's embeddings run locally with sentence-transformers — no API key or cost.

LLM tooling moves fast. If an API or package has changed since these versions, the architecture and patterns still apply — adjust the syntax to what you have installed.