EvanThomasLuke/Awesome-AI-Hacking-Agents

List of AI Hacking Agents

What it is

Awesome-AI-Hacking-Agents is a community-curated list ("awesome list") of AI agents used for hacking and security testing. It currently tracks roughly 64 open-source AI hacking agents in a table, plus a few research papers with no code.

Core premise

The README is explicit that these tools are for educational and authorized security testing purposes only. The maintainer is actively asking for contributions — you can submit missing projects via a GitHub issue.

What's in the list

The main content is a large table of projects, sorted by release date, from early 2021 (AutoPentest-DRL) through late 2025. Each row includes:

  • Name and GitHub/arXiv link
  • Creation/publication date
  • Notes — e.g., a USENIX Security 2024 paper for PentestGPT, an arXiv paper reference, or performance claims (Cyber-AutoAgent shows an "85% score on xbow top open source score")
  • CodeWiki and DeepWiki links — auto-generated documentation pages for each repo (some are marked as pending)

Project types vary: autonomous pentesting, vulnerability scanning, brute-forcing, SOC (security operations center) assistants, exploit automation, honeypots, code auditing, and more. The maintainer notes that "some technically aren't agents" and plans to re-sort and re-categorize.

Notable extras in the README

  • DeepWiki MCP integration tip: explains how to plug a free public MCP server (mcp.deepwiki.com/mcp) into AI IDEs (Cursor, Claude Code, Windsurf, Continue.dev) to query open-source repos — e.g., to inspect what features other agents have. Links to Devin's docs.
  • A Discord invite for discussing AI hacking agents.
  • A short to-do list: add last-commit dates and star counts in CSV form, add all benchmark scores, and merge Tencent repos.

Guardrails / caveats

The maintainer explicitly flags quality issues:

  • It's a work in progress, with entries missing links and comments still being added.
  • One entry (Shannon) carries a critical note that its claimed high benchmark score came from a white-box analysis on a black-box benchmark, which the maintainer calls "not a correct way to compare."
  • The table is machine-managed — there's a marker saying the table is handled by a script (awesome_table.py) and rows shouldn't be hand-edited.

Bottom line

A living directory of AI security agents — useful as a starting point to discover open-source tools, see which research papers they come from, and jump into project docs via DeepWiki/CodeWiki. Treat benchmark claims with skepticism (the maintainer does).

Related

  • Project
    microsoft/teams-sdk

    A comprehensive development framework for building AI-powered agents and applications integrated into Microsoft Teams.

  • Project
    docwriter-org/plain-writing-skill

    A set of plain-text rules and instructions that enable AI agents to write in a clear, concise style and remove unnecessary complexity from their responses.

  • Project
    Maneek21/Deft

    A self-hostable, open-source workspace that integrates chat, task management, and knowledge base tools with AI agents, featuring governed approval workflows and support for the Model Context Protocol.

  • Project
    myYangyunfan/dsh_desktop

    DSH Desktop is a native Windows/macOS client that bundles DeepSeek Harness (dsh) with a self‑contained Node runtime and 29 official plugins. It offers zero‑dependency, one‑click start, session management, balance display, and robust auto‑update/auto‑heal features. Since v0.5.0 it runs on a Tauri 2 (Rust) shell, providing a sleek frameless window, side‑session floating windows, and built‑in plugin manager. Installers are available for Windows, macOS and Linux; a portable zip is also provided. Licensed under MIT.