Day 1 — Foundations of Agent AI with the Claude Agent SDK
Goals for the day
- Build a shared mental model of what an Agent AI is and isn’t
- Get everyone’s environment up and running with the Claude Agent SDK
- Ship a first working agent that uses tools, has memory, and runs in a loop
Agenda
| Time | Session | Format |
|---|---|---|
| 09:30 – 10:00 | Welcome, intros, environment check | Round-the-room |
| 10:00 – 11:00 | What is an agent? Agent vs assistant vs workflow | Lecture + discussion |
| 11:00 – 11:15 | Break | — |
| 11:15 – 12:30 | Claude Agent SDK — anatomy of an agent (loop, tools, memory) | Lecture + live demo |
| 12:30 – 13:30 | Lunch | — |
| 13:30 – 15:00 | Lab 1 — your first agent: tool use + multi-turn | Hands-on |
| 15:00 – 15:15 | Break | — |
| 15:15 – 16:30 | The tool surface — built-in tools, custom tools, MCP overview | Lecture + demo |
| 16:30 – 17:00 | Day 1 retro, Day 2 preview | Round-the-room |
What participants will leave with
- A running local agent that uses at least one custom tool
- A clear mental model of the agent loop (model → tool → observation → model)
- An opinion on when not to use an agent (the most underrated lesson)
Materials
- Slide deck: link TBD
- Lab 1 starter repo: link TBD
- See References for SDK docs
Attendees
- NBS: TBD
- MetaWeave: TBD
Decisions
- Captured live during the session.
Action items
- Owner — action — due