# Zero to Agent Swarm, Part 2: A Team of Agents

This is the second part of the Zero-to-Agent-Swarm tutorial. In the first part, we went from zero to a working AI agent. If you want to check that out, it’s here: ← Part 1: Birth and Upgrades — building a single agent from scratch. This part is about going from one agent to an agent swarm — making agents work together for us. We’ll need a new mental model. In Part 1, the model was about what an agent is — Triggers + loop(LLM + Tools + Memory). Now we need to think about what agents do when there are many of them. We need to consider them — I hate to say this — more like humans. Not because AI is sentient, no, not by a large measure. But because we’ve built computers to resemble human functions, and we have centuries of experience making human systems productive. We can steal a lot of that for agent systems. Here’s the progression: Specialization — same code, different agents Delegation — agents calling agents Workspace — shared state for coordination Project execution — structured plan