I wrote a bit about my ClawPatrol setup for my agents on Tailscale: littledivy.com/clawpatrol
The firewall rules are very powerful since they are enforced at the network protocol level
Prevent prompt injection, DROP TABLES and do Human-in-the-loop without modifying the agent
Deno@deno_land
Introducing Claw Patrol! A security firewall for AI agents. It sits between your agents and production, parsing and gating HTTP, SQL, Kubernetes, and SSH on the wire. Built from our experience giving ops agents production access without sharing credentials. MIT licensed.
English
