0xSteph/pentest-ai

Open-source AI pentester that proves every finding. Machine oracles re-run each exploit; verified bugs ship a proof capsule you can replay yourself.

What it solves

Pentest-AI addresses the problem of "AI slop" in security testing—where AI models suggest vulnerabilities that are actually false positives. It provides a framework to connect AI models to real penetration testing workflows, ensuring that every reported finding is independently verified by a machine oracle before it is marked as a verified verdict.

How it works

The system uses an AI model to drive the investigation of web applications, probing for weaknesses like injection and access control issues. When a potential vulnerability is found, a dedicated "machine oracle" repeats the exploit and compares the result against a control to prove the vulnerability exists. This evidence is stored in a portable "proof capsule" that can be replayed to reproduce the finding. The tool can be integrated via MCP (Model Context Protocol) into AI clients like Claude Code or Cursor, run via CLI, or integrated into CI/CD pipelines.

Who it’s for

It is designed for authorized penetration testers and bug bounty hunters who need to automate the vulnerability discovery process while maintaining high signal and strict adherence to program scopes (e.g., HackerOne, Bugcrowd).

Highlights

  • Verified Findings: Uses independent oracles to prove exploits work, preventing the model from awarding its own verdicts.
  • Scope Enforcement: Strictly enforces program boundaries (wildcards, CIDRs) to ensure no out-of-scope targets are hit.
  • Bug Bounty Integration: Imports scope files from major platforms and generates platform-ready submissions including CVSS/CWE and reproducible curl commands.
  • Flexible Deployment: Supports local models via Ollama, cloud providers, and MCP integration for AI-native IDEs.
  • CI/CD Ready: Can be configured to fail builds based on verified findings and export results in SARIF format.

Related

  • Project
    s0ld13rr/pentestcode

    PentestCode is an open‑source, terminal‑based AI agent that automates full penetration‑testing workflows. A lead LLM coordinates 13 specialist sub‑agents (recon, scanner, enumerator, exploiter, etc.) that share a persistent structured engagement state, enabling the system to scan, enumerate, spray credentials, exploit, and report with minimal human prompting.

  • Project
    PentesterFlow/agent

    PentesterFlow is an open‑source terminal assistant that lets a human analyst guide an LLM‑driven agent through penetration‑testing tasks. It integrates real security tools (curl, shell, Burp, file utilities), supports many LLM backends, stores evidence and findings as markdown, and retains project‑level memory across sessions while requiring explicit approval for risky actions.

  • Project
    psyray/oasis

    OASIS is a Python CLI tool that uses locally hosted LLMs (via Ollama or any OpenAI‑compatible server) to scan source code for security vulnerabilities. It runs a fast scan model, then deep‑analysis models, validates findings deterministically, caches results, and outputs canonical JSON plus HTML/PDF/Markdown reports. A password‑protected web dashboard lets users explore findings and ask the assistant for re‑investigations. It supports multi‑model runs, suppression registries, diffing against baselines, and CI exit‑code gating.

  • Project
    Vasco0x4/AIDA

    AIDA is an open‑source autonomous pentesting platform that lets any LLM (Claude, Gemini, OpenAI, etc.) drive a Docker‑based security‑testing environment. The agent can run reconnaissance, scanning, custom Python exploits, and raw HTTP manipulations, logging every command and its output in a persistent notebook. Findings are auto‑scored with CVSS 4.0 and can be exported as PDF reports. The project includes a web dashboard, JWT authentication, notification hooks, and supports swapping the built‑in `aida-pentest` container for the larger Exegol image. Real CVEs discovered by AIDA are listed, showing practical impact. Licensed under AGPL‑v3.

  • Project
    GreyDGL/PentestGPT

    PentestGPT is an MIT‑licensed Python CLI that chains LLMs (Claude Code, Codex, or many other providers) into an autonomous penetration‑testing pipeline. It runs staged phases (recon → exploit → walkthrough for CTF, or asset discovery → vulnerability identification → report for real pentests), can persist sessions, and offers a legacy interactive mode where a user guides three cooperating LLMs. Install with `make install` (Python 3.12+, uv) or via Docker, then run `pentestgpt --target <IP>` (or `pentestgpt-legacy` for the interactive version). The tool reports optional anonymous telemetry and achieved ~86 % success on a published benchmark.