About PentestGPT
PentestGPT is an open-source AI-powered autonomous penetration testing agent, published at USENIX Security 2024. It uses a multi-stage pipeline that feeds findings from one phase into the next: for CTF challenges it progresses through reconnaissance, exploitation, and walkthrough; for penetration tests it goes through asset discovery, vulnerability identification, and report generation. The agent can operate autonomously using Claude Code or Codex, or interactively in legacy mode supporting a wide range of LLMs including OpenAI, Anthropic, Google Gemini, DeepSeek, xAI, Qwen, Moonshot, and local Ollama. Features include AI-powered challenge solving, real-time live walkthrough, multi-category support (Web, Crypto, Reversing, Forensics, PWN, Privilege Escalation), session persistence, and an extensible modular architecture.
Key Features
Pros & Cons
- Open source and free to use
- Published at USENIX Security 2024 (academic recognition)
- Multi-stage pipeline ensures thorough and structured testing
- Supports a wide variety of LLMs via legacy mode
- Session persistence enables long-running assessments without losing progress
- Modular design allows easy extension and customization
- Requires Python 3.12+ and uv package manager for installation
- Autonomous mode requires additional CLI tools (Claude Code or Codex) to be installed and authenticated
- Setup complexity may be high for non-technical users
- Interactive legacy mode is separate from the autonomous agent pipeline