Agentic WeeklyGet the weekly

ARTEX: an open-source multi-agent LLM system for autonomous penetration testing in isolated labs

Release Ruler for this brief: marked at its publication date, Oct 26, 2026, among the AI models we track.

ARTEX is an AGPL-3.0 open-source system in which a planner agent and worker agents, driven by Anthropic or OpenAI models, carry out penetration tests in an isolated lab, sharing an asset graph and an exploration graph, with human-in-the-loop approval and full traffic recording; it ships as a single Go binary or a Docker Compose stack.

By Agentic Weekly · Evidence: Reported · How we work

First seen
(via GitHub new-repository discovery sweep (topic ai-agent); primary source is the project's GitHub repository)
Source published
(we saw it 17.8 days later)
Brief published

What changed

  • The README describes a multi-agent penetration testing system for an isolated lab: a planner and workers share an asset graph and an exploration graph, with human-in-the-loop approval and full recording of traffic.
  • It has a Go backend and a Next.js frontend, needs PostgreSQL, and uses an ANTHROPIC_API_KEY or OPENAI_API_KEY (or a model set in the UI) for exploration. Install routes are an install.sh script, Docker Compose, or a single binary built from source.
  • The web UI shown in the README covers a dashboard with token usage, tasks and agent sessions with tool calls, an exploration graph, findings, assets, traffic recording, chat, interception approvals, agent and LLM settings, and remote MCP servers over HTTP or SSE.
  • Assets can be imported from the ScopeSentry asset-discovery tool via its API key, and the settings page can check for and install new versions with SHA256 checks and automatic rollback after three failed starts, per the README.
  • The repository is licensed AGPL-3.0, was created on 2026-10-08 UTC and had 1,970 GitHub stars when we saw it; its latest GitHub release is v0.3.15, published 2026-10-18 UTC.

In the source’s words

“AI autonomous penetration testing system. Multi-agent LLM, Go backend, Next.js frontend.”

“It ships as one binary or a Docker Compose stack.”

What we have not verified

  • We did not install or run ARTEX; every feature listed comes from the project's README and GitHub metadata.
  • The repository description says the project won a Baidu agent attack-and-defense challenge; we did not find or check a primary source for that claim.
  • We did not assess how well the agents find vulnerabilities or how safely they stay in scope. Penetration testing tools should only be pointed at systems you are authorised to test.
  • The source publish time shown is the repository's creation date; we first saw it on 2026-10-26 through our new-repository sweep.

This brief is Reported: written from the sources below, not from running the release ourselves.

Sources

Spotted an error? See our corrections policy or email the address on the contact page. Markdown version for agents: /news/artex-multi-agent-ai-penetration-testing.md

Want these briefs in your Google Top Stories? Add Agentic Weekly as a preferred source on Google.