The control plane for AI agents.

Govern what your AI coding agents do — not just what they say. Runtime policy enforcement for Claude Code, Cursor, Copilot, and more.

Process & file ACLs

Block AI agents from running scripts or touching files outside policy. Linux BPF_LSM, macOS EndpointSecurity, Windows in-process hooks.

Fail-closed by design

If the policy daemon goes down, agents stop. No silent passthrough. Built for endpoint security, not LLM output filtering.

Sandbox · Bumpers · Learning

Three policy archetypes covering the spectrum from strict allow-list to default-allow with deny rules to observe-only mode.