AI Runtime Assurance

Priced on the evidence, not the seats.

Start free and sign your first governed action today. Pay when the evidence has to outlive the week.

Start free. Pay when an agent goes into production.

Every plan signs every governed action as it happens. What changes is how much of it we run for you, how far back you can answer, and whether a third party can check the record.

Self-serve

Starter

Free

No card. Mint your first receipt in minutes.

  • 1,000 signed receipts a month
  • 30-day receipt retention
  • Console, live event stream, in-browser verification
  • Self-hosted SDK free forever, self-signed receipts
  • Pauses after 30 idle days, one click to wake
Most teams start here

Production

Team

$995/ month, billed annually

14-day trial, no card. Month to month is $1,195.

  • A dedicated arbiter, signing from the first operation
  • 1,000,000 receipts a month, then $5 per 10,000
  • 12-month retention, so last quarter is still answerable
  • Evidence packs for security reviews and audits
  • Independent witnessing with inclusion proof
  • Guided onboarding
Start 14-day trial

Then $995 a month, billed annually

Regulated scale

Enterprise

Talk to us

Annual agreement, from $36,000.

  • Everything in Team
  • Attested-blind enclave path, in co-design with a state AI program — built so we cannot read payloads
  • Private connectivity and deployment in your own environment
  • Retention set to your record-keeping obligation
  • Named support, SLA, and help through customer security reviews

Building at a startup?

Preferred pricing by application — $99/mo billed annually, or $129 month to month. Under $5M raised or under two years old.

Apply

The meter is signed receipts, not seats and not agents. You are billed for the thing the Console proves, which makes the invoice as verifiable as the product.

Evidence receipts

Signed proof that runtime controls executed.

Runtime controls execute locally and generate signed receipts written to the OVERT 1.1 open standard. Receipts chain, and third parties can verify control execution without seeing the sensitive payload.

  • OVERT-conformant signed receipts, assembled into evidence packs
  • Framework exports: NIST AI RMF, ISO 42001, EU AI Act, OSCAL
  • Zero sensitive-data egress: evidence packs assembled from local receipts
Verify a receipt

Pricing FAQ

Is the free tier really free?

Yes, and there is no card. The hosted free tier covers 1,000 signed receipts a month with 30 days of retention — enough to integrate, prove it works, and show a colleague. If you want it unmetered, run the self-hosted SDK: free forever, no cap, no retention limit, because when you run it yourself it costs us nothing to let you. A free workspace parks itself after 30 idle days and wakes on one click.

Why does the free tier only keep evidence for 30 days?

Because retention is the honest place to draw the line. Signing an action is cheap; keeping the proof available for the day somebody finally asks is the part that costs money and the part that carries the value. If a decision from last quarter has to be answerable, that is a paid plan.

What counts as a receipt?

One governed action. Not a token, not a request, not a seat. You are metered on the exact thing the Console shows you, so the invoice can be checked against the evidence the same way a customer checks your receipts.

What happens if we go over the included count?

Overage runs at $5 per 10,000 receipts on Team. Nothing stops signing and nothing gets dropped, because a gap in the record is the one failure mode this product exists to prevent. We talk to you before anything unusual reaches an invoice.

Is there a startup rate?

Yes, by application: $129 a month, or $99 a month billed annually up front, for companies under $5M raised or under two years old with an AI product in market. Participants in partner programs and state AI sandboxes qualify automatically. A person reviews every application, typically within one business day. Apply here.

We already use Vanta/Drata. Do we need GLACIS?

Vanta and Drata document that the policies exist. Glacis proves they were followed at runtime. They’re complementary. Together they close the AI evidence gap.

Sign your first governed action today.

Start free in minutes, or talk to us about the workflow your customers keep asking about.