Anthropic · CCAR-F
Multi-agent systems and the coordinator
CCAR-F · Lesson 2 of 18 · 9:35
YouTube loads only after you press play. This chapter opens at its timestamp in the full course. Watch on YouTube ↗
In this lesson
Prepare for the Claude Certified Architect – Foundations exam, on the exam guide version 1.0, effective July 2026. In this lesson: the coordinator-subagent pattern, hub-and-spoke architecture, isolated subagent context, decomposition, delegation, and aggregation, choosing subagents dynamically, partitioning the scope, decompositions that are too narrow, iterative refinement, and routing everything through the coordinator. Includes three original practice questions.
Preview the topics
- Kestrel’s coordinator sends every question through all four subagents, even a simple lookup of a single fact. What should change?
- A report on the European camping market covers only Germany and France, in depth. The coordinator’s log shows two subtasks, one per country, and both subagents succeeded. What is the root cause?
- A developer proposes letting the search and analysis subagents exchange findings directly, to save a round trip. What is the main cost of this change?