In active development
Cortex Strike
Autonomous Offensive Security Co-Pilot & VAPT Agent
Cortex Strike turns penetration testing into a conversation. You describe a target; an AI agent plans the assessment, chooses the right tools from a large offensive-security library, runs them in sequence, and reasons over the results — recon feeding exploitation feeding follow-up, without a human wiring each step by hand.
It runs on a locally-hosted open-source model, so nothing about a target ever leaves the machine. Around the agent sits a full product: authentication, persistent chat history, per-user tool selection, and a live view of each tool as it executes.
//what makes it work
Long tool-chains without losing the thread
Multi-step assessments generate enormous tool output that quickly overwhelms a model's memory. Cortex Strike keeps full results in external storage and gives the agent an on-demand way to look them up, so it can run long chains and still quote exact findings accurately instead of hallucinating them.
Built on a hardened foundation
Rather than start from zero, it stands on a substantially re-architected and hardened open-source security toolkit — refactored for modularity, stability, and safe deployment.
A real product, not a script
Auth, saved conversations, per-user tool personalization, and streamed live execution progress make it something a security team could actually sit down and use.
built with
- Python
- FastAPI
- React / TypeScript
- Local LLM
- PostgreSQL