See how DVARA governs model, tool and agent traffic. Thirty minutes, on your own use case.
A DVARA demo is a live walkthrough of one governed request end to end — who called, what policy decided, what the inspection stages did, and what was recorded — then the controls that matter to your case and how you would deploy it. Tell us what you are solving and we skip the generic tour.
What you enter on step one stays in this page — it is not submitted anywhere on its own and is not added to the calendar link. Booking is handled by our scheduling provider, which collects whatever its own form asks for and passes it to DVARA so we can prepare for the call. Analytics records that you moved between steps, and none of your answers. See the privacy notice. Whichever of these your case turns on. Almost all of it runs without a licence key — routing, policy, guardrails, PII, budgets, caching and the signed audit trail. A licence adds governance for tool calls and agent hops, which is where the walkthrough usually goes next. YAML rules on model, tokens, tools, MCP server and arguments, residency, time of day and budget use — dry-run against a request you describe, shadow against real traffic, then promote. Injection, jailbreak and output-sanitization patterns, plus PII detection over sixteen checksum-validated types with more available. Your workspace setting decides whether a match blocks, redacts or is recorded. Calls and the decisions taken on them are recorded, HMAC-signed and hash-linked so alteration and removal are detectable. Framework reports are assembled from those records as evidence for an assessment. Tool calls routed through it get policy on the arguments, PII scanning, approval gates on the tools you name, and loop detection. A licence activates it as a path on the gateway you already run — there is no second service to deploy. Agent-to-agent hops get caller identity, a registry of peers you allow, per-hop policy, inspection and approval — on their own audit chain, in their own store. Completed model calls priced from provider usage and your own rates, attributed per workspace and key, with budget caps that refuse calls and a soft breach that can downgrade the model. One OpenAI-compatible endpoint in front of fourteen first-class providers, with capability filtering before selection and failover to a healthy capable peer. Per-workspace policies, budgets and provider credentials, resolved workspace-first with a vault option and a strict mode that refuses the platform fallback. Isolation is enforced in the application layer. Prefer to try it first? The gateway and console boot from public images with no licence key and no account. Thirty minutes, live. We trace a request through the gateway — identity, the policy decision, what the inspection stages did, and what was recorded — then spend the rest on whichever of those matters to you. If you tell us what you are trying to solve beforehand, that is the part we go deep on rather than a tour of everything. So the half hour goes on your case. The answers decide which planes we show and which deployment shape we talk through — they are not a qualification gate, and you can skip straight to the calendar. Nothing is required except a time. No. Nothing needs installing for the walkthrough. It helps to know which providers you call and what is driving the work — an audit, a customer security review, a cost problem, or agents about to reach production — because that is what decides where the time goes. Yes. The gateway and console boot from public images with no licence key and no account, which is enough to run your own traffic through the OpenAI-compatible endpoint and see what it records. The governance engines need a licence, and that is usually the reason to book. Whoever owns the request path and whoever owns the obligation — in practice a platform or engineering lead alongside a security, compliance or FinOps stakeholder. The questions that decide a governance purchase come from both sides, and answering them separately takes twice as long. The answers on the first step stay in the page and are used to shape the conversation; they are not sent anywhere on their own, and they are not put into the calendar link. When you book, the scheduler collects whatever its own form asks for and that goes to DVARA so we can prepare and turn up. See the privacy notice for how we handle it.What does the walkthrough cover?
Common questions about the demo
How long is the walkthrough and what happens on it?
Why the questions before the calendar?
Do I need anything set up beforehand?
Can I try DVARA without booking anything?
Who should be on the call?
What happens to what I enter?