Cerberus

Cerberus provides an AI‑driven penetration testing cursor that enforces a mathematically proven safety boundary. Users specify a scope and a natural‑language objective, and the system generates a machine‑checked proof for each action, ensuring that any code outside the defined scope cannot execute. The tool runs locally, supporting on‑premise and air‑gapped environments, and can work with custom agents, VSCode extensions, and CI/CD pipelines.

The platform is aimed at security teams and enterprises that need continuous, production‑level testing without exposing credentials or data to uncontrolled AI behavior. It offers real‑time monitoring, collaborative workspaces, and automated client‑ready reports generated the same day. By leveraging type‑theory‑based restrictions, Cerberus claims to prevent destructive or out‑of‑scope actions that other AI pentesters cannot guarantee.

Distinctively, Cerberus combines formal safety guarantees with practical integrations such as SAST/DAST proof‑of‑concept generation, local model support, and the ability to operate in highly restricted environments. This approach is intended to reduce reliance on manual pentests while maintaining strict control over AI‑driven testing.