One loop, built by hand.

A finished, runnable micro-agent — the kind you can demo, not a toy snippet. The course walks through the agent loop, then layers on planning, memory, collaboration, and task isolation until the example genuinely orchestrates tasks.

No. Endpoints are simulated inside the app, so the agent runs against local previews that behave like real services — you learn the patterns without wiring infrastructure. Lesson files are export-first: take the Python files out and adapt them to a real stack later.

A course shaped like an app. You read structured lessons, then watch and step through the runnable example — the demo is the lab. It is built for learning how agents work, not for shipping production bots.

The loop is the heartbeat of every agent: observe, decide, act, repeat. Planning and memory are improvements on that loop; orchestration is many loops cooperating. Start with the simplest version and every later concept has somewhere to attach.

Developers and students who have read about agents but never opened one up. If "tool use" and "context" feel like buzzwords, building the loop with your own hands turns them into code you understand.