Robots Atlas>ROBOTS ATLAS
ARTEX-Pro
Active

ARTEX-Pro

ARTEX-Pro is an open-source, autonomous agentic system for penetration testing (AI autonomous penetration testing). It pairs a Go backend with a Next.js web interface and is model-agnostic — it can use any large language model (LLM) as the agent's "brain" to plan and run offensive security testing. The project is self-hosted and released under the AGPL-3.0 license.

ARTEX-Pro is an independently developed, modified variant based on the source of the original ARTEX project (author: Autumn-27). The wpsec/ARTEX-Pro repository independently maintains the code, Docker Compose configuration, container images and releases; it is not the original project's official channel. According to the repository, version 0.3.18 continues from native ARTEX 0.3.15. The base ARTEX is described as a separate entity.

The interface includes a dashboard (overview, token usage, activity stream), a task list and an execution-process view (sessions and tool calls), an exploration-chain graph, a findings list, assets together with a coverage graph in a force-directed layout (highlighting tested nodes, collapsing and expanding), traffic recording, human-in-the-loop chat, agent management, LLM configuration, interception and approval of actions (with an approval record) and backend logs.

Deployment is via Docker Compose, which by default starts ARTEX-Pro together with a PostgreSQL database (the default interface address is http://localhost:8787). OWASP ZAP runs as a process inside the ARTEX-Pro container and exposes web application testing tools through a built-in MCP server. The original project has been discontinued; the wpsec repository maintains the sources, Compose configuration, images and releases.

According to the repository's description, the project is the champion ("冠军项目") of Baidu's "agent+" challenge (攻防挑战赛 — an offensive-defensive challenge). The tool is dual-use (offensive security) and is intended for authorized penetration testing conducted only with the consent of the owner of the tested system.

Self-hosted (lokalnie)Web appCLIFree
Updated 17 Nov 2026

Overview

System type
Input
TextURLs
Output
ReportsRisk alertsGraph visualisationNetwork mapsCodeText

Access & Pricing

Access channels
Self-hosted (lokalnie)
Access by self-hosting on your own hardware — no vendor cloud, fully local.
Web app
Access through a browser-based web application.
CLI
Access via command-line interface.
Subscription model
Free

Open-source software (AGPL-3.0), self-hosted — no charge for the system itself. The real cost comes from the chosen LLM (API or your own inference infrastructure) and compute resources.

Capabilities

11 capabilities
Reasoning & Planning
Agency & Action
Task Specialization

Integrations

System tools

Security & Enterprise

Self-hosted — data stays within the operator's infrastructure. Agent behavior is constrained by approval gates and action interception (human-in-the-loop), built-in command-interception rules, scope rules and budgets; commands can be reviewed before execution.

Sources

2 sources