CrabTrap: An LLM-as-a-judge HTTP Proxy to Protect Agents in Production

CrabTrap is an open-source LLM-as-a-judge HTTP proxy that supports safely deploying AI agents in production environments.

This tool intercepts every HTTP request generated by an agent, checks it against configured security policies, and allows or blocks the request in real time. Judgment is made through two paths as follows.

• Static rule matching
• LLM judgment-based inspection

Users can directly add or modify new rules to finely control and monitor the agent's behavior.