Lesson 5 · 9 min
Agent State & Memory
Persisting decisions and intermediate results across turns.
Agents that span turns need durable state: which tools have been called, what was learned, what the user has approved. Persist a minimal state vector (JSON) rather than serializing the full conversation. Reconstruct context from the state vector on each turn.
Quick check
Agentic ArchitectureSelect one
Which action ALWAYS requires explicit human confirmation in a Claude agent?
