#1616th of 25 in Security

vulnclaw

LLM agent CLI that runs penetration tests from natural-language prompts

Stars
3.9k
License
MIT
Last commit
Oct 2026
Last release
Oct 2026
Language
Python
Stars, last 29 weeks+3,902
25 Apr 2026: 39 stars2 May 2026: 46 stars9 May 2026: 52 stars16 May 2026: 62 stars23 May 2026: 85 stars30 May 2026: 109 stars6 Jun 2026: 309 stars13 Jun 2026: 488 stars20 Jun 2026: 561 stars27 Jun 2026: 789 stars4 Jul 2026: 1,826 stars11 Jul 2026: 1,992 stars18 Jul 2026: 2,091 stars25 Jul 2026: 2,185 stars1 Aug 2026: 2,387 stars8 Aug 2026: 2,612 stars15 Aug 2026: 2,779 stars22 Aug 2026: 2,909 stars29 Aug 2026: 3,013 stars5 Sept 2026: 3,213 stars12 Sept 2026: 3,296 stars19 Sept 2026: 3,352 stars26 Sept 2026: 3,439 stars3 Oct 2026: 3,474 stars10 Oct 2026: 3,519 stars17 Oct 2026: 3,733 stars24 Oct 2026: 3,798 stars31 Oct 2026: 3,852 stars7 Nov 2026: 3,909 stars14 Nov 2026: 3,941 stars

39 on 25 Apr 20263,941 on 14 Nov 2026

From GitHub's daily star counts.

Checked on GitHub.

Overview

VulnClaw is a Python CLI and agent that takes a target and a plain-language goal, then lets the model drive recon, vulnerability discovery, exploitation and report generation. It uses built-in fetch, shell, Python and nmap tools plus MCP servers (Burp, Chrome DevTools). Tool output is stored as raw evidence, and claimed flags must match it. It also ships a REPL, a TUI, a local Web UI and a source-code scanner.

Who it is for: Pentesters and CTF players running authorized tests with an LLM agent

Strengths

  • Evidence gate rejects flags or conclusions that do not appear verbatim in tool output
  • 13 LLM provider presets including local Ollama, plus any OpenAI-compatible base URL
  • Generates Markdown reports and runnable Python PoC scripts after a run
  • Ships 50 skill reference packs the model loads on demand, plus a local code scanner

Weaknesses

  • Full Web UI requires building the React frontend from source; the PyPI wheel lacks it
  • python_execute is described as high-risk and not a strong sandbox
  • Burp and Chrome DevTools MCP integrations need separate manual deployment
  • README is mainly Chinese; the English version is in a separate file

What it needs

  • no GPU
  • Docker + Compose
  • Needs Python 3.10+, LLM API key or Ollama, Node.js 18+ (Web UI build, npx MCP servers), nmap (optional), Burp Suite Professional (optional), Chrome (optional)
  • Models: OpenAI, Anthropic, MiniMax, DeepSeek, Zhipu
  • port 7788

What changed

From the release notes. No flag means the note does not say so, not that the upgrade is safe.

Also in Security

See all 25
Also in Security
RankProjectScore
1shannonAutonomous pentester that exploits web apps and APIs using source analysis52k starsAGPL-3.092 out of 100
2ifixaiPython CLI that audits AI agents for policy, access and reliability failures46.3k starsApache-2.086 out of 100
3StrixAutonomous multi-agent pentesting CLI that validates vulnerabilities with working PoCs76.4k stars+7,005 (+10%) in 30 daysApache-2.085 out of 100
#4ai-infra-guardRed-teaming platform for AI infra CVEs, MCP servers, agent skills and jailbreaks7.5k starsApache-2.080 out of 100
#5failproofaiPolicy enforcement and run tracing for coding agent CLIs and gateways9.6k stars+3,964 (+70%) in 30 dayscustom license79 out of 100