METATRON: Offline AI Pentesting, Zero Cloud, Zero Compromise

amy 15/05/2026

What is METATRON?

METATRON is a local-first, CLI-based penetration testing assistant for Parrot OS. Runs nmap, nikto, whois + local LLM analysis—no API keys, no cloud, no subscriptions.

You point METATRON at a target. It runs real recon tools. Feeds results to a local AI model. Gets back vulnerability analysis, exploit ideas, and remediation steps. Everything stays on your machine. No telemetry. No paywall.

Built for Parrot OS. Runs offline. Powered by Ollama + a fine-tuned Qwen model. Stores every scan in MariaDB. Exports to PDF or HTML.

Built by Soorya Thejas!

How it works (in 4 steps)

  1. Target input: Drop an IP or domain into the CLI menu
  2. Automated recon: METATRON fires off nmap, whois, whatweb, curl, dig, nikto—no stubs, no mocks
  3. Local AI analysis: Results go to metatron-qwen (running via Ollama). The model correlates findings, flags risks, suggests next steps, and can even request additional tool runs mid-session
  4. Store + export: Full scan history lands in MariaDB. Edit, delete, or export clean reports in PDF/HTML: ready for clients or internal docs

Why it stands out

  • 100% local: No cloud calls, no API keys, no usage tracking. Your data never leaves the box.
  • Agentic workflow: The AI doesn’t just summarize; it can ask for more scans, dig deeper, and adapt its approach.
  • Real tools, real output: Not a wrapper around fake results. METATRON runs the same CLI utilities you’d use manually.
  • Structured history: Five linked MariaDB tables let you query past scans, compare targets, or audit findings.
  • Report-ready exports: One-click PDF/HTML output without leaving the terminal.
  • Free forever: MIT-style ethos. No feature gates. No “pro” tier.

Quick specs

Component Detail
OS Parrot OS (Debian-based)
Language Python 3
LLM Runner Ollama
Model metatron-qwen (fine-tuned Qwen 3.5, 9B/4B options)
Context Window 16K tokens
Database MariaDB (local)
Recon Tools nmap, whois, whatweb, curl, dig, nikto
Search DuckDuckGo (no key) + CVE lookup
Export PDF, HTML

Get it running (fast)

# Clone + setup
git clone https://github.com/sooryathejas/METATRON.git
cd METATRON
python3 -m venv venv && source venv/bin/activate
pip install -r requirements.txt

# System deps
sudo apt install nmap whois whatweb curl dnsutils nikto

# AI backend
curl -fsSL https://ollama.com/install.sh | sh
ollama pull huihui_ai/qwen3.5-abliterated:9b  # or :4b for <8GB RAM
ollama create metatron-qwen -f Modelfile

# Launch
python main.py

💡 Tip: The 4B variant runs comfortably on 8GB RAM systems. The 9B model prefers 16GB+.

Disclaimer

This tool is intended for educational purposes and authorized penetration testing only.

  • Only use Metatron on systems you own or have explicit written permission to test.
  • Unauthorized scanning or exploitation of systems is illegal.
  • The author is not responsible for any misuse of this tool.

Bottom line

If you do security work and want AI assistance that respects air-gapped workflows, client confidentiality, or just a clean offline setup, METATRON gives you smart analysis without sending a single byte to the cloud.

Run local. Test hard. Stay private.

Related articles