An AI pentester that logs in like a human first

What it does ARTEX runs LLM-driven agents that autonomously explore and test a target’s attack surface, recording sessions, tool calls, traffic, and findings along the way. It’s a Go backend with an embedded Next.js frontend, backed by PostgreSQL, and it ships as a single binary or Docker stack. Everything the agents do — traffic, findings, assets — is searchable and linkable in the UI, and there’s a human-in-the-loop chat plus interception approvals so the operator can step in before anything risky happens.

The interesting bit The bundled browser extension is the clever part: you log into an authenticated target yourself (captchas, email OTP, 2FA and all), and your browsing becomes agent-visible evidence — the captured requests land in the traffic store like proxy recordings, ready to bind to findings. It also syncs assets from ScopeSentry so you don’t re-collect what you already scanned.

Key highlights

• LLM-agnostic: Anthropic, OpenAI, or a keyless local Ollama server, configurable in the UI
• Force-directed asset coverage graph showing what’s been tested, with node collapse/expand
• Interception approvals and global approval records with expandable detail views
• “Restart equals migrate”: the binary idempotently re-runs schema.sql on every start, no manual migrations
• Ships with common pentest tools (nmap, curl, ripgrep, etc.) baked into the Docker image

Caveats

• This is a fork with translated docs — the original is by Autumn-27, and the repo history was rewritten in October 2026, so older clones need a manual reset
• No published Docker image; you build from checkout, and the previous default image pointed at a namespace that no longer hosts it
• Requires PostgreSQL and an LLM key (or local Ollama) — it won’t do much without both

Verdict Worth a look if you run authorized pentests and want an agent that works your asset graph while you supervise. If you wanted a fully hands-off scanner, the interception-approval workflow will feel like friction — by design.