CLUSTER PENETRATION-VULNERABILITIES-AGENT · 16 ENTRIES · EARLIEST ENTRY 2026-09-02
Autonomous AI penetration testers
This cluster comprises AI agents that autonomously perform penetration testing, including reconnaissance, exploitation, and reporting of vulnerabilities with proof-of-concept evidence. The dated history before the index start is too thin to read a shape from, as only one dated entry exists; it likely predates the index or its members are undated. Among traction leaders, DeepAudit leads with 6855 stars, followed by ARTIFEX with 4 stars and RedBee with 2 stars.
Entry timeline
ENTRY TIMELINE
Each bar is the month a member carries, which is a creation date where the source published one and a crawl date otherwise. Nothing in the index was found before 11 August 2026, and coverage broadened sharply on 14–15 August, so a bar from then on reports how much we found that month, not how much was built; an earlier bar reports how far back a member's own source dated it — not how long we have been watching.
Members by activity
Activity signals, not revenue.
Build origin · 1 ai-built · 15 unlabelled
Operator · 6 hobby · 2 indie · 1 commercial · 3 open source · 4 unlabelled
Two labels per entry, never collapsed into one: how it was built, and who stands behind it. Counted over the 16 listed below. Unlabelled means nothing has decided yet — not that the answer is small.
- 01DeepAudit6.9k★1 mo—open sourceMulti-agent AI system that automatically scans code for vulnerabilities, validates findings in a sandbox, and generates reports.
- 02ARTIFEX14★1 day—hobbyAutonomous penetration-testing system that uses multiple AI agents to identify and exploit security vulnerabilities in authorized systems.
- 03RedBee2★14 days—hobbyAI penetration-testing agent that automatically reconnoiters, exploits, and reports vulnerabilities with proof-of-concept evidence while learning from past engagements.
- 04Lares-Windows3★22 days—hobbyAutonomous security agent that hardens Windows machines by applying security benchmarks, verifying fixes, and rolling back changes that cause harm.
- 05grepleaks1★27 daysliveindieAI pentest agent that runs in the terminal, guiding security engagements and producing reports.
- 06BreachLabs1★23 daysliveai-builtindieAutonomous AI security engineer that scans, tests, verifies, and fixes software vulnerabilities inside the development loop.
- 07Tanli1★23 days—Autonomous red-team agent that plans multi-step attack paths against authorized targets, runs probes, and produces structured security reports with reproduction steps.
- 08Prozync-AI1★27 dayslivecommercialAI assistant for ethical hacking and cybersecurity that performs penetration testing, vulnerability scanning, and remediation in a cloud sandbox.
- 09bughunter-harness2★1 mo—Autonomous security agent that scans websites for vulnerabilities through a multi-stage pipeline and sends reports by email or Telegram.
- 10pentestcode-bin0★1 moliveopen sourceAI-powered terminal agent that autonomously runs penetration tests and security assessments for developers.
- 11PentestGPT0★6 daysliveopen sourceAI agent that autonomously performs penetration testing on computer systems and reports exploitable security vulnerabilities.
- 12hive-agent-security0★24 dayslivehobbySecurity tool that monitors populations of AI agents for dangerous emergent capabilities and automatically contains unauthorized data exfiltration.
- 13Master-peon0★1 mo—Containerized platform that pairs a conversational AI agent with dozens of penetration-testing tools to run authorized security assessments from natural-language commands.
- 14api-qa-skill0★24 days—hobbyAgent skill that performs live-probed, self-healing API test automation with defensive assertions and Allure reporting.
- 15randori-pentest-agent1★1 mo—hobbyAutonomous security agent that scans networks, detects vulnerabilities, exploits targets, and generates penetration test reports.
- 16hivepro-ai-associate0★1 mo—AI assistant that prioritizes software vulnerabilities using asset context and threat intelligence for security teams.
Ranked by traction, highest first, and paged 20 at a time — this is a window on the cluster, never the whole of it. Ties are broken by a stable id so a product cannot appear on two pages or fall between them. Browse every cluster.