FAQ

Common questions about UnitOne Gateway: blocking unsafe tool use, CRA-ready evidence, spend by agent, AWS Azure or GCP in-network deploy, and how to start.

What is UnitOne?

UnitOne Gateway inspects agent traffic: it stops unsafe tool use, keeps exportable CRA-style logs, and shows spend by agent, team, or project.

How is this different from a cost dashboard?

Cost dashboards show model spend. Gateway starts by blocking unsafe agent actions and keeping logs you can export for CRA-style duties, then attributes spend so you can set caps.

Does Gateway run in our network?

Yes. Gateway deploys in your AWS, Azure, or GCP network on Free, Team, and Enterprise. Agent traffic stays in your environment. See plans for agent, seat, and event limits. Agents are AI workloads, not human seats.

How does multi-cloud Gateway deploy work?

UnitOne Gateway runs in your AWS VPC, Azure VNet, or GCP project as the data plane: prompts, virtual keys, and spend stay with you. The UnitOne control plane holds tenant config, policy, inventory, and CRA evidence. Sync crosses the trust boundary over outbound HTTPS only — admin API, identity federation, and webhooks. See the homepage architecture section.

Is remediation included when I start free?

No. Finding → Fix Spec → PR work is a scoped proof of concept, not the self-serve product. Request a remediation POC if you need that engagement.

How do we get started?

Start free on Gateway (Stripe collects a card at checkout; Free is $0), see plans for Team at $99/month, or talk to us for Enterprise. Use Request a remediation POC only when you need a scoped fix-to-PR engagement.

What is EU CRA Article 14 for security teams?

Article 14 is the manufacturer reporting clock for actively exploited vulnerabilities and severe incidents. An SBOM is inventory, not that report. See the CRA Article 14 checklist for the operational evidence list.

How do you attribute spend to agents, not seats?

Agents are AI workloads through Gateway; seats are console users. Discovery, experiments, and published agents need separate identities and caps. See Agent discovery to publishing spend.