Human-in-the-loop (HITL)
A checkpoint where a person must approve, edit or reject an agent's action before it proceeds.
HITL inserts a human decision into the agent loop — typically before side-effecting actions (deploy, send, pay, delete). It is the most common control in production: LangChain's survey found most teams require human approval for write/delete tools. The design question is where to place checkpoints so that they catch errors without making the agent slower than doing the task manually.