Open-Source ARTEX Turns AI Agents Into Autonomous Hacking Teams
ARTEX is an open-source multi-agent system that autonomously runs penetration tests end-to-end using a dual-graph planner-worker architecture.
- ARTEX is an open-source autonomous pentesting system in Go with a Next.js UI, released under AGPL-3.0
- Dual-graph design separates per-task exploration from a globally shared asset graph, linked by anchors
- Planner dispatches intents, N worker agents claim and execute them with real tools through a MITM proxy
- Workers can search each other's execution transcripts to avoid duplicate work on parallel branches
- Supports Anthropic, OpenAI, and local Ollama models; one-liner Docker install at github.com/BaaRai/ARTEX
- Ships a browser extension for authenticated traffic capture and ScopeSentry asset sync
ARTEX, an open-source autonomous penetration-testing system, is now available as an English translation of Autumn-27’s original Chinese-language project. The repository packages a Go backend, an embedded Next.js interface, PostgreSQL, and a swarm of LLM-powered agents that run tools such as Bash, HTTP clients, Nmap, ripgrep, and curl against a defined scope. At the time shown, it had attracted 320 stars and 173 forks.
ARTEX provides an inspectable implementation of a planner-worker architecture for offensive-security research. Its companion norma SDK supplies the agent core, tool interfaces, permissions, memory, and execution transcripts, while a recording man-in-the-middle proxy captures HTTP requests and responses as shared evidence. Operators run the system locally and must restrict it to infrastructure they own or have explicit authorization to test.
Two graphs separate targets from tactics
ARTEX divides state between a global Asset Graph and a task-specific Exploration Graph, preventing infrastructure data from becoming tangled with an agent’s decisions and observations.
| Graph | Scope | Stored data |
|---|---|---|
| Asset Graph | Shared across tasks | Root domains, subdomains, IP addresses, services, applications, and endpoints |
| Exploration Graph | One per task | Goals, intents, facts, findings, hints, and their lineage |
Anchors connect exploration activity to the affected assets. An operator can inspect which intents tested a particular service or trace a finding through relationships such as
This story is for Pro members
You've reached the end of the free preview. Upgrade to AlphaSignal Pro to read the full article - and everything else behind the paywall.
Comments
