Skip to content

Chapter 10: Agent Teams — Multi-Agent Orchestration

This chapter examines when plural workflows help: subagents that investigate in separate context, worktrees that separate tracked-file lanes, and roles that split bounded ownership. Parallelism adds coordination and shared-resource risk, so the chapter treats one agent or sequential work as valid outcomes.

This is the summit of the technical stack, the material the $500 cohorts save for their finale — and every piece of it stands on things you already own: context discipline (5), scoped briefs (2.3), verification you don't have to remember (9).

What you will learn

  • Delegate to subagents: isolated context, specialized roles, clean reports back.
  • Use Git worktrees to separate tracked-file lanes, then manage integration and shared-resource conflicts.
  • Orchestrate agent teams on real features: split, brief, integrate, review.
  • Assemble the full agentic harness — the machine that makes all of it trustworthy.
  • Operate as the orchestrator: allocation, review altitude, and knowing when one agent is plenty.

Builder principle

You've been the director since Chapter 1. This chapter just gives you a bigger crew — same standards, same receipts, more hands.

Watch and read with this chapter

Real talk

There's a moment in every builder's multi-agent week where it stops feeling like a feature and starts feeling like a staff. Three terminals, three workstreams, you moving between them like a foreman with a clipboard — approving here, redirecting there. The moment usually arrives with a laugh: this used to be a team's output. It still is. The team is just... you, multiplied. That's not hype; that's this chapter, practiced.

Lessons in this chapter

Chapter checkpoint

Complete Chapter 10: Parallel only when justified, save the evidence, then score your first attempt with the shared rubric.