Products · ARTEX
ARTEX is an autonomous, LLM-driven penetration-testing platform. A planner decides what to look at next; several workers run real tools against the target; everything they find flows back into one shared picture of the system. A human watches, chats, and steers at any point. One binary, on your own infrastructure.
Not a scanner with a checklist. It is a small team that reasons, packaged as one binary.
ARTEX runs a penetration test the way a small team would. A planner reads the situation and emits concrete directions. Several workers each claim one direction, run real tools through a Kali-style shell and HTTP, and write back what they find. An event-driven engine closes the loop: every change to the graph wakes the planner, which emits new directions, until the goals are proven.
The planner decides what to investigate from what is already known. It does not walk a list: it opens new directions only when something is uncovered, and stops when nothing is.
It separates what the target is (domains, services, endpoints) from how far we have probed it (goals, facts, findings). The two graphs cross-reference through anchors, and real coverage falls out of that.
An operator watches the work live, chats with the system, and steers it: add hints, set goals, mark what is out of scope. You do not wait for a report to intervene.
A finding is a confirmed, reproducible vulnerability with its traffic stored. A reporter agent turns it into a write-up the board and the technical team can both read without a translator.
A single Go binary with the frontend embedded, backed by PostgreSQL. It deploys on your own infrastructure: the target's traffic and the evidence do not leave for a third-party SaaS.
Each task carries a cumulative LLM token budget. Spend is measured and capped per task, so a reasoning loop never becomes a surprise invoice.
The heart of ARTEX: what the target is, and how it was reasoned about.
The exploration graph is the lineage of the reasoning. A goal spawns intents; each intent, run by a worker, yields facts and findings; a fact can derive a fresh intent. Not a checklist — the chain of decisions, with its evidence anchored to real assets.
Real screens from an engagement — against a lab target, with synthetic data.
Real screens from the ARTEX interface, in English. Demonstration target (acme.com) and synthetic data.
ARTEX drives real tools against real systems. Everything it does is fenced, recorded and approvable.
Every task registers an authorized asset scope. Before a direction is claimed, authorization on those assets is re-checked; a block rule forces a close. Nothing gets touched by accident.
A recording MITM proxy captures every request and response, and authorizes egress. No agent action without a trace: a finding's evidence is the traffic that produced it, not a screenshot.
Sensitive actions pass through a gate: the operator can require manual approval before a tool runs. The system is autonomous by default and haltable in an instant.
A tool that really attacks only goes to those who are really accountable.
ARTEX drives real offensive tools against real systems. That is why we do not sell it to anyone. We license it exclusively to publicly accountable organizations — listed companies and public-sector bodies — able to issue a formal, documented authorization for every system under test.
Companies listed on a regulated market, and public-sector bodies. These are entities with governance, disclosure duties and a formal process to authorize a test — not an anonymous email address.
Every engagement starts from a signed, authorized scope: which systems, for how long, under which limits. That scope is what ARTEX enforces in code, not a line in a contract.
Pointing ARTEX at something you neither own nor are permitted to test is not a supported use — it is an abuse of the product. The asset scope and the approval gate exist to make that line technical, not only moral.
We do not hand over a binary and leave. ARTEX lands with deployment on your infrastructure, integration with your identity and telemetry, a runbook, and the same operating contract as everything else we do.
If you are not sure your organization fits these criteria, the answer is probably not yet — and we are happy to say so plainly. Access is deliberately narrow.
The same things we already run in production — not a wish list.
If you are a listed company or a public-sector body and want to test your own attack surface continuously, send us two lines about your organization and the scope you have in mind. Within 24h you get the next steps and the authorization requirements.