PentAGI

PentAGI (Penetration testing Artificial General Intelligence) is an open-source, self-hosted platform for automated security testing aimed at information-security professionals, researchers, and enthusiasts. It runs an AI-powered agent that autonomously determines and executes penetration-testing steps, with optional execution monitoring and intelligent task planning for reliability. All operations happen inside a fully isolated, sandboxed Docker environment, and the platform bundles more than 20 professional security tools such as nmap, metasploit, and sqlmap. Rather than a single monolithic agent, PentAGI uses a team of specialists with a delegation system: an orchestrator coordinates researcher, developer, and executor agents that share discoveries and chain findings. It has a smart long-term memory system for storing successful approaches, plus a Graphiti-powered knowledge graph on Neo4j for semantic relationship tracking. Built-in web intelligence includes an isolated browser scraper and integrations with external search systems such as Tavily, Traversaal, Perplexity, DuckDuckGo, Google Custom Search, Sploitus, and Searxng. For builders and teams, PentAGI ships a clean web UI plus full REST and GraphQL APIs secured with Bearer tokens for automation and integration. It supports 10+ LLM providers including OpenAI, Anthropic, Google AI/Gemini, AWS Bedrock, Ollama, DeepSeek, GLM, Kimi, and Qwen, plus aggregators like OpenRouter and DeepInfra, and custom OpenAI-compatible endpoints. Observability is first-class, with Langfuse for LLM analytics and Grafana/Prometheus, Jaeger, and Loki for monitoring. Data persists in PostgreSQL with the pgvector extension, and the microservices architecture supports horizontal scaling. It generates detailed vulnerability reports with exploitation guides and is deployed quickly via Docker Compose. Note it is an autonomous and assistant-guided pentesting platform, not a breach-and-attack-simulation product. Use only on systems you are authorized to test.