Relay: metering and billing
What Relay charges for, what it never charges for, how the request meter reaches your Google Cloud invoice, and how private offers work.
Two lines, never three
Relay's price has two components and only two:
- Plan fee — annual, per organisation, billed through Google Cloud Marketplace as a subscription (prorated by the second if you start mid-month, per Google's billing engine).
- Requests — one unit per relayed model call beyond the plan's monthly allowance, reported hourly to Google Cloud and shown as a metered line on the same invoice.
There is no model-token line on our invoice. Relay only operates on your own provider keys, so tokens are billed to you by the provider at your own price. This is deliberate: a gateway that resells tokens has an interest in your volume; Relay does not.
The request meter
| What | Metered |
|---|---|
| Completed model call, any wire format, streaming or not | 1 request |
| Refused by allow-list, budget or guardrail block | 0 |
| Provider error after retries (5xx, timeout) | 0 |
| Guardrail checks, trace, dashboards, export | included |
Allowances are per calendar month and do not carry over. Usage is netted against the allowance before anything is reported; a month inside the allowance reports zero.
Every reported unit is backed by a trace with a request_id; the console's Billing → Reconcile page lists the exact requests behind each reported hour so your finance team can tie the Marketplace line to evidence.
How it reaches the invoice
After purchase on Marketplace, Google notifies Relay; the organisation is linked to the purchase and the plan's environments and allowances are provisioned. Usage is reported through Google's Service Control API in the metric gateway_requests (unit: 1 request). Your billing administrator sees the subscription and the metered line under Marketplace in the Cloud Billing console, and the charge counts toward Google Cloud committed spend.
Private offers
Enterprise terms — private deployment, longer retention, SLA, data residency, multi-year — are sold as private offers: we send a quote to your Google Cloud billing account ID, you accept in the console, and the terms replace the public plan for your organisation. Private offers follow the public structure (plan + environments + included requests); discounts are volume or term based.
AWS
An AWS Marketplace listing with the same two dimensions (contract + request consumption) will follow. Until then AWS-only customers can be invoiced directly.
Changes
Prices are locked for the term of your plan. Allowance or meter definition changes are announced in the changelog at least sixty days before they apply to renewals.