Claw Patrol is an open-source security proxy for AI agents. It sits between your agent and the network, injects credentials the agent never sees, and enforces HCL approval rules — with humans or LLM judges in the loop for risky actions.