Use AgentClick when an autonomous agent needs a human-in-the-loop review step before execution. The agent can open a local browser UI for command approval, plan review, code diff review, email or draft editing, trajectory inspection, selection capture, and memory review, then continue with the human’s edits or decision. The boundary is narrow and publishable: this is a browser review layer for agent actions, not a generic web app listing and not a broad agent platform card. Invoke it when terminal-only approval is too thin and a richer review checkpoint is the actual missing workflow.