I've been building with LLMs for a while, and security has been a constant headache. Manual red teaming is slow and doesn't scale, so I built a tool to automate it.
Agent Aegis is an autonomous system that stress-tests your LLM apps. It uses a team of specialized AI agents that work together to:
Profile your AI to understand its function and personality.
Generate & run tailored attacks, from simple prompt injections to complex jailbreaks.
Judge the responses, score vulnerabilities, and give you actionable steps to fix them.
The goal is to make robust AI security testing accessible to everyone, not just big teams.
The stack is React/TypeScript/Tailwind on the front end, with the Gemini API powering the agent logic.
It's still early days, and I'd love to get your feedback, especially on the multi-agent architecture and the effectiveness of the generated attacks.
Agent Aegis is an autonomous system that stress-tests your LLM apps. It uses a team of specialized AI agents that work together to:
Profile your AI to understand its function and personality.
Generate & run tailored attacks, from simple prompt injections to complex jailbreaks.
Judge the responses, score vulnerabilities, and give you actionable steps to fix them.
The goal is to make robust AI security testing accessible to everyone, not just big teams.
The stack is React/TypeScript/Tailwind on the front end, with the Gemini API powering the agent logic.
It's still early days, and I'd love to get your feedback, especially on the multi-agent architecture and the effectiveness of the generated attacks.
You can try it here:
Thanks!