ARTEX autonomous penetration testing system released as open source
Understand this story
BaaRai has open-sourced ARTEX, a self-hosted, LLM-driven multi-agent penetration testing platform. The planner is responsible for breaking down objectives, workers invoke real tools to execute, and results are written back to two linked graphs: the asset graph and the exploration graph; each finding can be bound to captured requests and responses as evidence. All autonomous operations must pass through scope rules, human approval, and token and turn budgets. Deployment requires PostgreSQL; the exploration feature requires configuring an LLM (ANTHROPIC_API_KEY or OPENAI_API_KEY, which can also be configured in the interface). There are no prebuilt Docker images; images must be built from local code; the interface listens on 127.0.0.1 by default. The untrusted output boundary, which is enabled by default, places tool results received by workers, the planner, chat, and the main agent within boundaries marked with random tokens, so that the model treats them as data rather than instructions.
AI overview based on source reports · Updated 2 hr ago
Coverage timeline
Follow the coverage to explore different perspectives.
- Hacker News AIArtex – AI autonomous penetration testing system
ARTEX is a self-hosted, LLM-driven penetration testing platform in which a planner decomposes goals into intents, workers execute them with real tools, and results are written back into two linked graphs (an asset graph and a per-task exploration graph) joined by anchors. Findings are bound to captured request/response evidence via a recording MITM proxy and browser extension, and autonomy is constrained by scope rules, approval gates, token and turn budgets, a noise governor, and prompt-injection framing of tool output. It ships as a single Go binary with an embedded Next.js UI, requires PostgreSQL, and supports Anthropic, OpenAI-compatible, or local Ollama endpoints.
Activity for this story
Not enough continuous observations to show a trend yet.