vxcontrol / pentagi

项目摘要

PentAGI is a fully autonomous multi-agent AI system purpose-built for automated penetration testing. It orchestrates AI agents within sandboxed Docker environments to execute complex security testing workflows while supporting multiple LLM providers and self-hosted deployment.

为什么值得关注

It uniquely combines multi-agent AI orchestration with offensive security in a privacy-focused, self-hosted platform that has rapidly gained 21k+ stars, featuring broad LLM provider support and active development with regular releases.

优势

• Multi-agent architecture decomposes complex penetration testing tasks into specialized agent workflows
• Sandboxed Docker isolation ensures safe execution of potentially dangerous security operations
• Extensive LLM provider compatibility (OpenAI, Anthropic, Ollama, Gemini, Bedrock, DeepSeek, GLM, Kimi, Qwen)
• Self-hosted deployment model gives security teams full data control and compliance flexibility

局限性

• Lacks automated tests despite being a security-critical tool, raising reliability concerns
• Go language ecosystem offers fewer AI/ML libraries compared to Python-based alternatives

使用场景

• Automated penetration testing and vulnerability assessment for enterprise security teams
• Red team operations with AI-assisted reconnaissance and exploitation planning
• Security research automation for discovering novel attack vectors and exploit chains