Autonomous penetration testing: the questions security leaders ask.

R0 by ReactiveZero is an on-prem autonomous penetration-testing appliance. A local model running on NVIDIA DGX Spark hardware you own orchestrates scoped subagents that drive real offensive tooling across web, API, cloud, hybrid identity and on-prem Active Directory, correlates findings into CVSS 3.1-scored attack chains and records every step in an append-only ledger. No vendor cloud; air-gap capable.

Autonomous penetration testing is a machine planning and executing a real penetration test end to end — choosing tools, chaining findings and deciding when to stop — rather than following a fixed scanner playbook. R0 does this with a local model on an appliance you own, so the depth of a run is driven by your estate, under a strict scope guard, not by an iteration limit. See how an agentic pentest works.

No. R0 is an on-prem appliance running a local model on your own hardware. Findings, evidence and model inference never leave the box. The only traffic that leaves your network is R0 reaching the in-scope targets you authorise — the same packets your own pentester would send. It can also run fully air-gapped for internal-only engagements.

R0 is non-destructive by default, with a strict scope guard, SSRF and anti-loop guards, an instant kill switch and an append-only, replayable ledger of every action. Higher-impact actions — credential changes, device enrolment — run only under the explicit rules of engagement, after an operator approval gate, and are reversible from the ledger.

R0 reports demonstrated attack chains, not scanner hits. A finding is only raised when a subagent actually reproduced it, and every finding ships with the steps to reproduce so your team can verify it. ReactiveZero consultants validate results before a report is delivered.

The SaaS tools rely on a vendor cloud control plane for orchestration and results. R0 has no vendor cloud to sync to — it runs entirely on hardware you own, which suits regulated and sovereign environments. It also chains external, cloud, hybrid-identity (Entra ID / AD Connect) and on-prem Active Directory attacks in a single autonomous run, rather than focusing on the web tier alone.

R0 is delivered as an appliance built on NVIDIA DGX Spark (GB10 Grace Blackwell). The local model is the brain; a Kali toolset in a Docker sandbox is the hands, isolated and reset between runs. More units mean more parallel subagents and deeper runs.

Model, template and tool updates are packaged by ReactiveZero and applied on your schedule. Nothing has to leave your network for R0 to run — you decide when and how an update package crosses the boundary.

A set of demonstrated, CVSS 3.1-scored attack chains with evidence and reproducible steps; an append-only ledger you can replay step by step; and a blue-team detection checkpoint for each step so the report sharpens your monitoring. Machine-readable exports are available for your tracker and SIEM. See Controls & audit.

R0 does the running; people own the judgement. ReactiveZero scopes the engagement with you, sets the rules of engagement, reviews and validates findings, and handles retests. You get autonomous coverage with a certified red team standing behind the results. See How to buy.

R0 is built and operated by ReactiveZero, an offensive-security consultancy in Rotterdam. Its consultants hold OSCP, CEH and CHFI certifications and carry out pentest and red-team engagements themselves. See About R0.

Pricing is on request; it depends on the appliance configuration and the scope of your engagement. Book a demo or join the on-prem waitlist and we will talk it through.