PentAGI
An autonomous multi-agent system for penetration testing that plans and runs security testing steps inside an isolated Docker sandbox, with specialist agents for research, development and infrastructure work.
PentAGI is a tool for automated security testing built around autonomous AI agents. It is aimed at information security professionals, researchers and enthusiasts who need a flexible system for conducting penetration tests, with all operations run inside an isolated, sandboxed Docker environment.
An AI-powered agent determines and executes penetration testing steps on its own, with optional execution monitoring and task planning, delegating work to a team of specialized agents for research, development and infrastructure tasks. It ships with more than 20 professional security tools such as nmap, Metasploit and sqlmap, a long-term memory system that stores research results and successful approaches for reuse, an optional Neo4j-backed knowledge graph for tracking semantic relationships, a built-in browser for gathering information from the web, and integration with external search APIs such as Tavily, Firecrawl and Perplexity.
It supports LLM providers including Ollama, OpenAI, Anthropic, Google AI, AWS Bedrock, DeepSeek, GLM, Kimi, Qwen and MiniMax, and offers optional Langfuse-based monitoring and OAuth integration. It is written in Go with a React frontend and licensed under the MIT license.