Open-Source METATRON Runs AI Hacking Scans Without Leaking Data to the Cloud
METATRON is a fully offline, open-source AI penetration testing assistant that runs a local LLM to automate recon, vulnerability analysis, and reporting with zero cloud exposure.
Takeaways
- Fully offline AI pentesting: METATRON runs a local LLM via Ollama to analyze recon data with zero cloud exposure.
- Agentic loop: The AI can autonomously request additional tool runs mid-analysis using structured
[TOOL:]and[SEARCH:]tags, up to 9 iterations. - Six recon tools integrated: nmap, nikto, whois, dig, whatweb, and curl -- all sandboxed via a strict allowlist to prevent arbitrary command execution.
- Full persistence and reporting: All findings are stored in a five-table MariaDB schema and exportable as PDF or HTML reports for client delivery.
- Hardware requirement: Minimum 8.4 GB RAM for the 9b model; a 4b variant is available for lower-spec machines.
- Free and MIT-licensed: No API keys, no subscriptions -- designed for air-gapped and compliance-sensitive environments.
METATRON is a new open-source penetration testing assistant that puts a locally running LLM in the driver's seat of your security workflow. You give it a target IP or domain, and it fires off real recon tools, feeds the results to an on-device AI model, and hands you back a structured vulnerability report -- all without a single byte of your scan data touching a cloud API.

The Problem It Solves
Most AI-assisted security tools today are cloud-dependent. You paste target data into a ChatGPT window or hit an API endpoint, and suddenly your internal IP ranges, service banners, and discovered vulnerabilities are sitting on someone else's servers. METATRON's most significant differentiator is its zero-exfiltration guarantee -- all LLM inference happens on-device through Ollama, meaning sensitive target data, including internal IP ranges, banner information, and discovered vulnerabilities, never leaves the tester's machine.
Running LLM inference locally changes the threat and compliance calculus for penetration testing. For red teams and security engineers, METATRON offers a way to leverage natural-language reasoning and LLM-driven command synthesis without sending sensitive environment metadata to cloud providers. That lowers data-exfiltration risk and may simplify compliance in regulated contexts. It also enables reproducible, air-gapped testing workflows where internet access is restricted or undesirable.
How It Works
METATRON is a CLI-based AI penetration testing assistant that runs entirely on your local machine -- no cloud, no API keys, no subscriptions. It runs real recon tools (nmap, whois, whatweb, curl, dig, nikto), feeds all results to a locally running AI model, and the AI analyzes the target, identifies vulnerabilities, suggests exploits, and recommends fixes.
The brain of the system is metatron-qwen -- a custom-configured variant of huihui_ai/qwen3.5-abliterated:9b, tuned specifically for penetration testing tasks and served locally via