pentagi

Why we included this project

PentAGI is a self-hosted system that lets an AI agent run a real penetration test from start to finish. It spins up a multi-agent setup inside an isolated Docker sandbox, so the LLM can drive tools like nmap, Metasploit, and sqlmap without touching the host. The agent decides its own next steps, and you can keep a human in the loop when you want to check its work. It remembers what it found across engagements, so techniques that worked before carry over, and it writes up vulnerabilities with exploitation guidance instead of just flagging them. If you are evaluating agentic security tooling, the wide LLM provider support, from OpenAI and Anthropic to Ollama and Gemini, makes it easy to test with whatever model you already use.