dhunter
An autonomous penetration-testing platform where an LLM agent plans and runs reconnaissance, active testing and vulnerability verification against an authorized target, then writes a report.
dhunter is an AI-driven, autonomous penetration-testing platform for authorized security research. Rather than only matching known CVE fingerprints like a vulnerability scanner, an LLM agent drives a curated set of tools to reconnaissance a target's subdomains, JavaScript, fingerprints and historical URLs, then plans attack intents that a shared blackboard coordinates across several parallel workers.
Every finding is verified before it is reported: a gate mechanically replays the proof of concept and rejects results that are not reproducible, cutting down false positives. Confirmed vulnerabilities are compiled into a Markdown report. The platform also has a library of reusable MCP tool connections, agent skills and scenario templates, and supports running several workers per project concurrently, pausing and resuming runs, and exporting a project's full findings.
It requires Go for the server and built-in tools and Python for the agent scheduler; several external scanners are optional. The material states it is for academic and security research only, on authorized targets.