ARTEX
Self-hosted, LLM-driven penetration testing platform where a planner and workers use real tools, and every finding stays tied to captured assets and HTTP traffic as evidence.
Visit BaaRai/ARTEXOverview
ARTEX is a self-hosted, LLM-driven penetration testing platform. A planner splits a goal into intents, workers execute them with real tools, and results are written back into two linked graphs: one for known target assets and one for test progress. Findings can stay bound to captured assets and HTTP traffic so evidence supports the result.
Key Features
- Dual-graph architecture separates the shared asset graph from each task's exploration graph and links them with anchors for coverage and audit trails.
- Evidence-bound findings use a recording MITM proxy and browser extension so findings, evidence quotes, and retests tie to exact request and response traffic.
- Autonomy controls include human-in-the-loop approvals, argv-prefix rules, scope and deny gates, token and turn budgets, a target noise governor, and prompt-injection framing on tool output.
- Self-hosted and model-agnostic deployment uses one Go binary with an embedded Next.js UI, PostgreSQL, and Anthropic, OpenAI-compatible, or local Ollama endpoints.
- Vulnerability retest loop re-runs findings with optional fixed version or conditions and records still reproducible, fixed, or cannot confirm.
- Reports, exports, task archives, asset coverage, and ScopeSentry sync support repeatable security engagements.
Use Cases
- Penetration testers and red teams that want automation without giving up an audit trail.
- Security teams that need to justify findings to clients or teammates with captured request and response evidence.
- Engagements against authenticated targets where a browser extension captures session traffic.
- Repeatable security engagements using task categories, templates, relations, archives, and asset coverage.
Getting Started
- Clone the repository, copy .env.example to .env, set POSTGRES_PASSWORD and optionally an LLM key, then run docker compose up -d --build.
- Open the UI on local port 8787 and complete first-run setup with the one-time setup token printed at startup.
- Retrieve the setup token from the artex container logs and complete the setup page.
Deployment & Requirements
- PostgreSQL is required; exploration requires an LLM such as an Anthropic or OpenAI-compatible endpoint, or a local Ollama instance.
- Docker Compose is one installation path; other paths include an install script, prebuilt binaries, source builds, and release archives.
- The UI listens on 127.0.0.1 by default; remote access requires binding or a reverse proxy or SSH tunnel.
- The Docker image includes a scanner toolchain such as nmap, nuclei, httpx, katana, ffuf, dalfox, sqlmap, ripgrep, curl, and npm.
Before You Adopt
- License: AGPL-3.0. Review its terms before using, modifying, or distributing the project.
- The Docker image is built from the local checkout, so there is no prebuilt image to pull and the first start can take longer.