ARTEX

ARTEX is an AI system for autonomous penetration testing in an isolated lab. A planner agent and worker agents share an asset graph and an exploration graph, so they build on each other's findings, while a human approves intercepted actions and every request is recorded.

The web UI covers a dashboard with token usage, task execution with sessions and tool calls, exploration chains, findings, an asset coverage graph, traffic recordings, approval records and agent and LLM configuration. Assets can be imported from ScopeSentry, remote MCP servers can be added, and the skills, agents and MCP setup are designed to be modified in forks.

It needs PostgreSQL and an Anthropic or OpenAI key, and installs through a script, Docker Compose or a single Go binary with the Next.js front-end embedded. In-app updates verify checksums, smoke-test the new binary and roll back after repeated failed starts. It is AGPL-3.0, its docs are largely in Chinese, and it is meant for authorized testing only.