PentAGI is a fully autonomous AI-powered penetration testing system built in Go that automates complex security testing tasks through multi-agent coordination. The project, maintained by vxcontrol, provides a self-hosted solution that combines artificial intelligence with professional security tooling to enable organizations to conduct comprehensive penetration tests without manual intervention.
The system operates as a multi-agent architecture where specialized AI agents handle distinct responsibilities including research, development, and infrastructure assessment. These agents coordinate through a delegation system and can be monitored during execution for enhanced reliability. The platform supports over 10 different large language model providers including OpenAI, Anthropic, Google AI/Gemini, AWS Bedrock, Ollama, DeepSeek, GLM, Kimi, and Qwen, as well as aggregators like OpenRouter and DeepInfra. This flexibility allows users to choose their preferred AI backbone or deploy local models for complete data sovereignty.
PentAGI includes a built-in suite of more than 20 professional penetration testing tools such as nmap, metasploit, and sqlmap, all executed within isolated Docker containers for security. The platform features a smart memory system that stores research results and successful approaches for future reference, complemented by a Neo4j-powered knowledge graph using Graphiti for semantic relationship tracking and advanced context understanding. Web intelligence capabilities are provided through an integrated browser scraper, while external search integration connects to multiple sources including Tavily, Traversaal, Perplexity, DuckDuckGo, Google Custom Search, Sploitus, and Searxng for comprehensive information gathering.
The architecture employs a microservices-based design supporting horizontal scaling, with all commands and outputs persisted in PostgreSQL enhanced with pgvector extensions. The system provides both REST and GraphQL APIs with Bearer token authentication, enabling programmatic access and integration with external systems. A modern web interface offers intuitive management and monitoring capabilities, while detailed logging integrates with Grafana and Prometheus for real-time system observation.
PentAGI generates thorough vulnerability reports with exploitation guides and supports intelligent task planning optimized for smaller models. The platform includes smart container management that automatically selects Docker images based on specific task requirements. Authentication is flexible, supporting GitHub and Google OAuth integration alongside API token systems. Deployment is streamlined through Docker Compose with comprehensive environment configuration options.
The chain summarization system manages conversation context growth by selectively summarizing older messages while preserving critical conversation flow, preventing token limit exhaustion. Configuration options allow customization of summarization behavior through environment variables controlling preservation of recent messages, QA pair strategies, and section size limits. Assistant instances can use customized summarization settings distinct from global configuration for fine-tuned context management.
The repository is classified as an AI security tool and offensive security platform, combining autonomous agents with penetration testing capabilities. It represents a departure from traditional breach and attack simulation products, focusing instead on autonomous and assistant-guided penetration testing workflows. The project emphasizes complete control over deployment and data through self-hosting while maintaining the flexibility to integrate with various external services and LLM providers.
