RadVector API clinical inference for radiology

Pricing

You pay for decisions, not for traffic.

Metered per clinical decision, with no seat license, no platform fee and no minimum. A call that returns nothing clinical costs nothing.

Current rate

RadVector API is metered per clinical decision: you are charged for decisions the engine reaches, and drafting an IMPRESSION around them is metered separately. There is no seat license and no minimum commitment. The live per-unit rates load here from our billing provider.

Read live from our billing provider, so this page and your invoice can never disagree. Usage is metered as it happens and reported on a schedule; your dashboard shows the month to date.

Try it before you decide

Create an account, mint a key, and make your first call in a couple of minutes. The API reference is public — you can read every route and schema before you sign up for anything.

What counts as a billable unit

A decision

One finding evaluated against the applicable criteria, returning a verdict with its provenance. Send three findings in one call and that is three decisions — the unit follows the clinical work, not the HTTP request, so batching costs you nothing and saves you round trips.

An impression

One drafted IMPRESSION from /v1/impression, priced separately because it is the one route that runs a language model. The verdicts in the same response are decisions as usual.

Not billed

Errors on our side, rate-limited calls, health checks, the API reference, and a call that reaches no decision. If the impression model is unavailable, you still get the verdicts and the impression is not charged.

Enterprise

For platform volume, a committed rate, a dedicated deployment, or terms your procurement team needs to sign, talk to us rather than to a pricing page. What is worth knowing before that conversation:

  • Built to need as little PHI as possible, with a BAA for clinical use. Identifiers have no field in the contract, free text is masked before the engine reads it, and request text is never logged or stored. Where calls will carry protected health information, including an AI or clinical platform calling RadVector on behalf of a hospital, we sign a Business Associate Agreement before they start.
  • Your calls are not training data. They are metered for billing and bounded for abuse, and request text is never logged.
  • Criteria versions are visible and stable. You can pin what a verdict cited, and see when guidance moved.

Questions a page cannot answer?

A radiologist will be on the call — this is built by the people who read the studies.