Inspiration
Penetration testing costs $15-50K per engagement. Solo developers, startups, and open source maintainers can't afford it. We asked: what if the Claude subscription you're already paying for could also be your security team?
Then we went further: what if two AIs fought each other over your code - one attacking, one defending and you could watch it happen in real-time?
What it does
Helios is an AI-powered security agent that autonomously finds, exploits, and fixes vulnerabilities. It features three modes:
Adversarial Agents Arena - Two Claude instances battle in real-time. Red Agent scans your code, crafts custom exploits, and attacks a Docker sandbox. Blue Agent monitors logs, patches vulnerabilities, and deploys WAF rules. Swarm mode lets you pit multiple attackers against one defender.
Supply Chain Attack Surface Analysis - Point Helios at multiple repos. It maps dependency relationships and identifies transitive attack vectors - vulnerabilities in one package that create exploitable paths in packages that depend on it. We tested this on Inkscape Web + django-cms and found 62 transitive attack vectors.
Autonomous Scanner - Fast local scanning with static analysis, secret detection, dependency checking, and configuration auditing. Finds SQL injection, XSS, SSRF, path traversal, command injection, hardcoded secrets, and misconfigurations.
Log in or sign up for Devpost to join the conversation.