Claw Patrol

Claw Patrol is an agent proxy and security firewall for AI agents. It guards credentials, intercepts traffic at the wire, and gates actions according to rules that the operator writes, while keeping an audit log of everything that happens. The page frames it as a way to sit between permissive access and fully unrestrained agent operation.

Its rule engine can match on HTTP method, path, headers, or body, as well as SQL verb, table, function name, and substrings of a statement. The same system also handles Kubernetes traffic, where rules can match namespace, resource, verb, and name. Examples shown on the page include blocking specific Postgres functions, checking message content with an LLM judge before HTTP requests go out, and requiring approval for Kubernetes exec activity. Claw Patrol also supports human-in-the-loop approval through Slack, the dashboard, or a webhook, and it can route requests through an LLM judge. The audit log and admin dashboard support drilling into requests and reviewing what the gateway captured.

It is meant to be used with any agent. The page shows agent commands prefixed with clawpatrol run, with the same workflow while actions are gated and tracked. Delivery is shown as a shell install script, and the documentation and demo are linked from the site. The page also says the tool can be extended with plugins.