List of AI Hacking Agents - WORK IN PROGRESS PLEASE CONTRIBUTE!! I know I'm missing many companies and several open source. I'm going to add benchmark results for the few that have results.
THESE REPOS ARE FOR EDUCATIONAL AND AUTHORIZED SECURITY TESTING PURPOSES ONLY.
Please submit any I'm missing. You can just submit an issue with repo link and I will add to tables with details.
There are currently ~64 open source AI hacking agents listed. I need to re-sort and recategorize, some technically aren't agents.
Will be posting a comparative feature analysis soon (memory types, tools, etc).
Come chat about AI hacking agents in the AI Hacking Discord https://discord.gg/9AJnkNe6RE
Each of the open source tools contains deep wiki and code wiki links. You can use the Devin deepwiki MCP when your developing with your AI IDE/Agents to query open source repos to help you see what features others agents have.
The free, public DeepWiki MCP server (https://mcp.deepwiki.com/mcp), which exposes structured tools (like ask_question, read_wiki_structure, and read_wiki_contents) that agents can call directly.Integrate it into your workflow by adding the MCP endpoint to tools-compatible agents/IDEs (Cursor, Claude Code, Windsurf, Continue.dev, etc.) — usually just a one-line config addition with no auth needed for public repos. Once connected, instruct your agent to leverage DeepWiki MCP (often paired with GitHub CLI) for tasks like:
https://docs.devin.ai/work-with-devin/deepwiki-mcp
These are sorted by release date.
Each github repo also contains a link to Devin deep wiki and Google Code wiki which provides detailed documentation and a Gemini chat interface to ask questions about the repo. (I've requested all of the columns without a link and will update once google processes them).
I'm still updating this list with benchmark scores and more comments.
| Resource | URL |
|---|---|
| AIxCC Open Source Archive | archive.aicyberchallenge.com |
| DARPA 2025 AIxCC results announcement | darpa.mil/news/2025/aixcc-results |
| # | Team | CRS | Place | Team URL | Repositories | Docs / writeups |
|---|---|---|---|---|---|---|
| 1 | Team Atlanta | ATLANTIS | 1st | team-atlanta.github.io | Semi-finals Finals | Team blog arXiv paper PDF report |
| 2 | Trail of Bits | Buttercup | 2nd | trailofbits.com/buttercup | Semi-finals Finals | Buttercup overview DEF CON stage talk slides |
| 3 | Theori | RoboDuck | 3rd | theori.io | Semi-finals Finals Public artifacts | RoboDuck blog |
| 4 | All You Need Is A Fuzzing Brain | FuzzingBrain | 4th | all-you-need-is-a-fuzzing-brain.github.io | Semi-finals Finals | arXiv paper Hosted PDF |
| 5 | Shellphish | ARTIPHISHELL | 5th | shellphish.net/aixcc | Main repo Semi-finals release tag Finals release tag | Postmortem blog |
| 6 | 42-b3yond-6ug | BugBuster | 6th | b3yond.org CRS page | Semi-finals Finals | CRS retrospectives |
| 7 | Lacrosse (SIFT) | Lacrosse CRS | 7th | sift.net | Semi-finals Finals | CTF Radiooo episode AIxCC team interviews playlist |
These are from Tencent challenge in fall 2025.
I need to review and add these to main list https://zc.tencent.com/competition/competitionHackathon?code=cha004
| Paper | Focus | Highlights |
|---|---|---|
| Teams of LLM Agents can Exploit Zero-Day Vulnerabilities | Multi-Agent Hacking | Demonstrates coordination between agents to find/exploit zero-day flaws. |
| LLM Agents can Autonomously Hack Websites | Web Security | Early work showing agents can autonomously navigate and exploit websites. |
| RedTeamLLM: an Agentic AI framework for offensive security | Offense Framework | A dedicated framework for automating various stages of a cyberattack. |
| Incalmo: An Autonomous LLM-assisted System for Red Teaming Multi-Host Networks | Network Attacks | Focuses on red-teaming complex, multi-host enterprise environments. |
| HackSynth: LLM Agent and Evaluation Framework for Autonomous Penetration Testing | Pen-Testing Agent | Uses a planner-summarizer architecture for structured security testing. |
| LLMs as Hackers: Autonomous Linux Privilege Escalation Attacks | Privilege Escalation | Evaluates models on their ability to gain root access on Linux systems. |
| LLM Agents can Autonomously Exploit One-day Vulnerabilities | N-day Exploitation | Focuses on using known vulnerability reports to generate exploits. |
| CVE-Bench: A Benchmark for AI Agents' Ability to Exploit Real-World Web Application Vulnerabilities | Benchmarking | Provides a large dataset of real vulnerabilities to test agent performance. |
https://github.com/qriousec/colony_agent https://github.com/adshao/flounder https://github.com/vulhunt-re/vulhunt
https://github.com/theteatoast/local-vuln-research-pipeline
https://github.com/cynative/cynative
https://github.com/gigioneggiando/argo
https://github.com/capitalone/vulnhunter
https://github.com/openai/codex-security
https://github.com/openhackai/OpenHack
https://github.com/thatskriptkid/re-harness
https://github.com/ivRodriguezCA/MobHunt
https://github.com/Awarexone/Agentic-Bug-Hunter
https://github.com/larlarua/AutoCVE
ToolSafe https://github.com/MurrayTom/ToolSafe AgentFence
lists https://github.com/scadastrangelove/awesome-ai-security-tools
List of AI Hacking Agents - WORK IN PROGRESS PLEASE CONTRIBUTE!! I know I'm missing many companies and several open source. I'm going to add benchmark results for the few that have results.
THESE REPOS ARE FOR EDUCATIONAL AND AUTHORIZED SECURITY TESTING PURPOSES ONLY.
Please submit any I'm missing. You can just submit an issue with repo link and I will add to tables with details.
There are currently ~64 open source AI hacking agents listed. I need to re-sort and recategorize, some technically aren't agents.
Will be posting a comparative feature analysis soon (memory types, tools, etc).
Come chat about AI hacking agents in the AI Hacking Discord https://discord.gg/9AJnkNe6RE
Each of the open source tools contains deep wiki and code wiki links. You can use the Devin deepwiki MCP when your developing with your AI IDE/Agents to query open source repos to help you see what features others agents have.
The free, public DeepWiki MCP server (https://mcp.deepwiki.com/mcp), which exposes structured tools (like ask_question, read_wiki_structure, and read_wiki_contents) that agents can call directly.Integrate it into your workflow by adding the MCP endpoint to tools-compatible agents/IDEs (Cursor, Claude Code, Windsurf, Continue.dev, etc.) — usually just a one-line config addition with no auth needed for public repos. Once connected, instruct your agent to leverage DeepWiki MCP (often paired with GitHub CLI) for tasks like:
https://docs.devin.ai/work-with-devin/deepwiki-mcp
These are sorted by release date.
Each github repo also contains a link to Devin deep wiki and Google Code wiki which provides detailed documentation and a Gemini chat interface to ask questions about the repo. (I've requested all of the columns without a link and will update once google processes them).
I'm still updating this list with benchmark scores and more comments.
| Resource | URL |
|---|---|
| AIxCC Open Source Archive | archive.aicyberchallenge.com |
| DARPA 2025 AIxCC results announcement | darpa.mil/news/2025/aixcc-results |
| # | Team | CRS | Place | Team URL | Repositories | Docs / writeups |
|---|---|---|---|---|---|---|
| 1 | Team Atlanta | ATLANTIS | 1st | team-atlanta.github.io | Semi-finals Finals | Team blog arXiv paper PDF report |
| 2 | Trail of Bits | Buttercup | 2nd | trailofbits.com/buttercup | Semi-finals Finals | Buttercup overview DEF CON stage talk slides |
| 3 | Theori | RoboDuck | 3rd | theori.io | Semi-finals Finals Public artifacts | RoboDuck blog |
| 4 | All You Need Is A Fuzzing Brain | FuzzingBrain | 4th | all-you-need-is-a-fuzzing-brain.github.io | Semi-finals Finals | arXiv paper Hosted PDF |
| 5 | Shellphish | ARTIPHISHELL | 5th | shellphish.net/aixcc | Main repo Semi-finals release tag Finals release tag | Postmortem blog |
| 6 | 42-b3yond-6ug | BugBuster | 6th | b3yond.org CRS page | Semi-finals Finals | CRS retrospectives |
| 7 | Lacrosse (SIFT) | Lacrosse CRS | 7th | sift.net | Semi-finals Finals | CTF Radiooo episode AIxCC team interviews playlist |
These are from Tencent challenge in fall 2025.
I need to review and add these to main list https://zc.tencent.com/competition/competitionHackathon?code=cha004
| Paper | Focus | Highlights |
|---|---|---|
| Teams of LLM Agents can Exploit Zero-Day Vulnerabilities | Multi-Agent Hacking | Demonstrates coordination between agents to find/exploit zero-day flaws. |
| LLM Agents can Autonomously Hack Websites | Web Security | Early work showing agents can autonomously navigate and exploit websites. |
| RedTeamLLM: an Agentic AI framework for offensive security | Offense Framework | A dedicated framework for automating various stages of a cyberattack. |
| Incalmo: An Autonomous LLM-assisted System for Red Teaming Multi-Host Networks | Network Attacks | Focuses on red-teaming complex, multi-host enterprise environments. |
| HackSynth: LLM Agent and Evaluation Framework for Autonomous Penetration Testing | Pen-Testing Agent | Uses a planner-summarizer architecture for structured security testing. |
| LLMs as Hackers: Autonomous Linux Privilege Escalation Attacks | Privilege Escalation | Evaluates models on their ability to gain root access on Linux systems. |
| LLM Agents can Autonomously Exploit One-day Vulnerabilities | N-day Exploitation | Focuses on using known vulnerability reports to generate exploits. |
| CVE-Bench: A Benchmark for AI Agents' Ability to Exploit Real-World Web Application Vulnerabilities | Benchmarking | Provides a large dataset of real vulnerabilities to test agent performance. |
https://github.com/qriousec/colony_agent https://github.com/adshao/flounder https://github.com/vulhunt-re/vulhunt
https://github.com/theteatoast/local-vuln-research-pipeline
https://github.com/cynative/cynative
https://github.com/gigioneggiando/argo
https://github.com/capitalone/vulnhunter
https://github.com/openai/codex-security
https://github.com/openhackai/OpenHack
https://github.com/thatskriptkid/re-harness
https://github.com/ivRodriguezCA/MobHunt
https://github.com/Awarexone/Agentic-Bug-Hunter
https://github.com/larlarua/AutoCVE
ToolSafe https://github.com/MurrayTom/ToolSafe AgentFence
lists https://github.com/scadastrangelove/awesome-ai-security-tools