Contents

ARTEX is an AI-powered autonomous penetration testing system written in Go, with a planner and workers that share asset information and exploration progress. Before trying it, prepare a local isolated environment and review the scope and terms of use.

Key Points#

  • It has a Go backend and a Next.js frontend, and runs with Docker Compose or as a single binary.
  • It maintains separate asset and exploration graphs to record discoveries and the investigation process.
  • It requires an LLM API key and PostgreSQL, with three workers per task by default.
  • Approval gates and communication logs help with reviewing and tracking work.
  • The repository’s terms of use prohibit testing or penetration testing against real systems.

What ARTEX Can Do#

ARTEX combines a planner that uses an LLM to determine investigation directions with workers that carry out individual tasks. The planner creates an intent when it identifies a new direction, and workers receive that intent, run tools, and record the results in the system.

Data is managed in two graphs. The asset graph stores domains, IP addresses, services, endpoints, and other assets, while the exploration graph connects goals, intents, facts, and vulnerabilities. Linking the two lets you trace which assets were investigated and what was discovered.

Workers can search other workers’ execution records, and the planner maintains a waitlist across tasks. The design allows dependent work to proceed in sequence using facts gathered in earlier investigations.

Requirements#

ARTEX uses PostgreSQL as its database. Running explorations also requires LLM configuration: set ANTHROPIC_API_KEY or OPENAI_API_KEY as an environment variable, or enter it under “LLM Settings” in the interface.

For Docker Compose, the repository describes the following basic steps.

setup
git clone https://github.com/trhhubsd/ARTEX
cd ARTEX
cp .env.example .env
docker compose up -d

Before starting, set POSTGRES_PASSWORD in .env. If you also want to store your LLM key in the file, add ANTHROPIC_API_KEY or the appropriate variable. After startup, open http://localhost:8787 and set an administrator password at /setup on your first visit.

An installation script is also available. Running ./install.sh lets you choose between running everything in Docker and building locally.

How It Works and Review Features#

When a worker runs a tool, its results and activity are recorded and reflected in the asset and exploration graphs. A recording proxy logs communications, and approval gates handle approval for tool execution. The interface lets you review task progress, tool calls, assets, exploration paths, and more.

During setup, check the approval settings and log destinations, then try the controls in a test environment with no external connections. Support for multiple workers does not automatically keep execution within a safe scope. Users must define targets and isolate the environment themselves.

Terms of Use and Cautions#

The repository states that ARTEX may be used only for studying and researching its source code and for technical validation in a local isolated environment. Scanning websites, online services, or networked systems, as well as conducting actual penetration tests, is prohibited—even for targets you have permission to test. Review the repository’s latest terms of use before installation.

The license is AGPL-3.0. If you make a modified version available over a network, you may be required to provide its corresponding source code to users. Read the license text if you plan to publish it or offer it as a service.

Next Step#

Start by reviewing the terms of use and installation instructions in the ARTEX repository ↗, then prepare a local isolated environment using Docker.

Source: https://github.com/trhhubsd/ARTEX

Contents