Agentic Pentester
Autonomous AI agent that plans, runs and reports web pentests. LangGraph planner, judge and summarizer driving real security tools.
- LangGraph
- FastAPI
- FFUF
- OWASP ZAP
- Playwright
- Python
- A LangGraph orchestrator that runs penetration tests end to end: a Planner agent picks the next attacks, tools execute them, a Judge agent rejects false positives, and a Summarizer writes up what survived.
- Drives FFUF for endpoint fuzzing, OWASP ZAP for spidering and scanning, and Playwright for JavaScript-heavy apps, with room for custom attack scripts.
- A confidence engine scores every finding with CWE mappings and severity, and an evidence store ties each one to the proof that produced it.
- Runs as a single-shot scan, an iterative plan-execute-learn loop until findings converge, or continuous monitoring, all configured in YAML with a max-iteration cap.
- Guardrails skip destructive actions and respect robots.txt; reports export as HTML, JSON or Markdown.