Model sources and pricing
Relay prices each model from the cheapest available source (Cloudflare Workers AI on Vibgrate’s account, or OpenRouter), then adds 10%. That margin reflects the token reduction from graph-grounded context — not a hidden markup on a single reseller.
When a model can run on Workers AI, that is the default serve path. The customer-facing model id stays stable; each source keeps its own underlying model id for routing.
Prepaid credits
Relay uses prepaid credit on the workspace.
Before a request is sent to a model, Relay estimates the cost and holds that amount against your balance. If the balance cannot cover the hold, the request fails with HTTP 402 and nothing is sent upstream. After the model responds, Relay settles to the actual token cost and returns any unused hold. You pay for tokens delivered, not for failed attempts.
No prepaid credit yet? Use BYOK (your own provider key) or local Code Modes with VG Code — both work without buying Relay credit first.
Enterprise customers on a signed contract can use an invoiced, post-pay arrangement with a contractual credit limit — ask your account team.
Buying credit
Top-ups are $25 / $50 / $100 / $250 / $500 / $1,000, by card, from dash.vibgrate.com → Settings → Billing → Credits. A processing fee is added on the receipt:
| Top-up | Processing fee |
|---|---|
| $100 and above | 3.9% |
| Under $100 | 4.9% |
Your card is charged credit + fee. The workspace receives the full top-up amount you selected (the fee is a separate line, not deducted from the credit).
Paid top-ups and subscription invoices appear under Settings → Billing → History (/settings/billing/history). Open a row for the Stripe PDF or receipt.
Auto top-up
Optional. Under Settings → Billing → Credits → Auto top-up, you can:
- Turn auto top-up on
- Set a threshold (e.g. $10)
- Choose a top-up size ($25–$1,000)
- Set a daily cap so a runaway agent cannot charge more than that per UTC day
When balance falls below the threshold after a request settles (or a request is refused for insufficient credit), Relay charges your saved card for one top-up. Credit is granted the same way as a manual purchase. A failed charge turns auto top-up off and notifies the workspace — it does not retry into a card decline. You need a payment method on file (Billing portal) before enabling.
Credit expiry
Each purchase is a lot that expires 12 months after purchase. Spend uses the lot closest to expiry first. You get email reminders 30 and 7 days before a lot expires.
Spend protection
Every request must reserve estimated cost first. On top of the prepaid balance:
- Monthly spend ceiling (optional) — set under Settings → Billing → Credits; a workspace at its ceiling cannot spend more that month.
- Per-request share — by default one request may reserve at most 50% of available balance (configurable).
- Auto top-up — daily cap and disable-on-failure (see above).
- Anomaly alert — if today’s spend is well above your trailing 7-day average, owners get email and an in-app notice (once per day).
- Credit expiry — lots expire 12 months after purchase; you are emailed at 30 days and 7 days before expiry.
Top up under Settings → Billing → Credits when you need more headroom.
How a request is billed (reserve → serve → settle)
The true cost of a streaming completion isn't known until the response ends, so Relay never serves first and bills after:
- Estimate — a conservative, rounds-up cost is estimated from your prompt size and requested max output tokens.
- Reserve — that estimate is atomically held against your balance. If the balance can't cover it, the request is refused before anything is sent upstream.
- Serve — the request is proxied to the model and streamed to you.
- Settle — once the provider reports actual usage, the reservation is corrected to the real cost and the unused hold is returned to your balance. A failed or errored request settles at zero — you are never charged for tokens you didn't receive.
Prompt caching
When a provider reports cache usage, Relay passes it through rather than billing at list price: cached reads are billed at the cheaper cache rate, and cache writes (which cost providers more) are billed at their real premium. When a provider reports no cache detail, tokens are billed at list — Relay never infers a cache hit it wasn't told about.
What you pay for
- Tokens the provider reports as consumed (including cache rates when the provider reports them).
- The processing fee on credit top-ups (shown on the receipt).
You are not charged for a request that errors before any tokens are produced, or for tokens beyond what the provider reports.