AITeamGuide Hub · Agent
PentAGI
A self-hosted system of autonomous agents for penetration testing, running tools in isolated containers with a web UI and reports.
Agent★ 32k starsArchived on GitHubMITSecurity
Key facts
- Type
- Agent
- Jobs it fits
- Pentest a network; Automate security testing
- Fields
- Security
- Works with
- Self-hosted
- GitHub
- vxcontrol/pentagi
- Licence
- MIT
- Stars
- 31,703 as of October 21, 2026
- Last updated
- October 17, 2026 (last push to GitHub)
- Author
- @vxcontrol
- Submitted by
- @aiteamguide
- Added to the Hub
- October 21, 2026
Only test systems you own or are authorised to test.
Install / Use it
MIT
Start (after configuring .env as in the README)
docker compose up -dSafety: Runs code from a third-party repo. Review the source and the permissions before installing.
Commands and configs are taken from the project's README and re-wrapped in each app's documented format. If they differ, the README wins.
Prefer a ready-made product?
Commercial agents for the same job, with real prices: Dropzone AI, all security operations agents.
More from the Hub
-
Skill
SkillSpector
NVIDIA's scanner that checks agent skills and MCP servers for malicious patterns, prompt injection and data leaks before you install them.
Security★ 21k
-
Skill
Trail of Bits Skills
Security research and code-audit skills from the security firm Trail of Bits.
Security · Software development★ 7.7k
-
Agent
Strix
Open-source AI agents that act like a penetration tester: they run your app, try real attacks and report vulnerabilities with proof and fixes.
Security · Software development★ 71k
-
Agent
promptfoo
A command-line tool to test prompts, agents and RAG apps, and red-team them for jailbreaks, data leaks and other LLM vulnerabilities.
Security · Software development★ 26k
Summary written by AITeamGuide in our own words. GitHub numbers come from the public GitHub API and refresh daily. Maintainer and want a correction or removal? Tell us. Got something similar? Submit it to the Hub.