Skip to content
Billing for AI products

Usage based billing built for AI.

Tokens, credits, requests, agent steps: meter what your product actually does and bill it precisely. Prices read the way model cards quote them, sub cent math is exact, and every charge carries a receipt your customers can open.

14-day free trial · everything on · no credit card

From $39/mo after.

Usage this month
$2.50 per 1,000,000 tokens
$600.00
240,000,000 tokens × $2.50 per 1,000,000
0 240,000,000 tokens 500M
Prefer prepaid? Sell a credit wallet instead: 1,000 credits = $9.00, burned down per request, topped up automatically.
Three ways to charge

Pick the model your product deserves.

Metered usage

Post usage events as they happen and bill in arrears: $2.50 per 1,000,000 tokens, graduated tiers, volume discounts, or an included free allowance. You type the price the way the model card quotes it; we handle the six zeros.

Prepaid credits

Sell wallets: 1,000 credits for $9.00. Usage burns the balance down per request, customers see credits and dollars side by side, and auto top up keeps them running. Low balance alerts included.

Passthrough with markup

Your app already knows what each request cost you. Push the exact amount, add a markup percentage, and yRecurring bills the sum. Your margin is a setting, not a spreadsheet.

Agents and beyond tokens

Bill agents by the step, the tool call, or the outcome.

Usage is anything you can count: agent steps, tool calls, image generations, audio seconds, GPU hours. Meter several dimensions on one subscription and mix them with a flat platform fee, seat pricing, or a percentage. It is one engine, so hybrid pricing is configuration, not a special case.

  • Safe ingestion: replays and retries never double bill
  • Sub cent rates handled in exact decimal, rounded once at the line total
  • Spend visibility for your customers: live balances and open receipts
  • Nothing extra to run: no event pipeline to deploy or babysit
Why it matters

Usage billing dies without trust.

The first surprise invoice is the last one an AI customer pays without questions. Every amount yRecurring produces carries a calculation trace: the meter reading, the rate, the math. Your customers can check the receipt instead of opening a dispute, and your support team answers billing emails with a link instead of a spreadsheet.

AI billing questions, answered.

How do I bill per million tokens?

Create a metered price and type it the way the model card quotes it: $2.50 per 1,000,000 tokens. Post usage events from your app and yRecurring aggregates them per customer per period, applies tiers or included allowances, and puts the math on the invoice with a trace.

Can I sell prepaid credits?

Yes. Define a pricing unit like credits with a fiat conversion (1,000 credits = $9.00), sell wallet top ups, and let usage burn the balance down automatically. Customers see credits and dollars side by side, get low balance alerts, and can auto top up.

Can I pass my model costs through with a markup?

Yes. If your app already computes the exact cost of each request, push that amount as a usage event and set a markup percentage. yRecurring bills the summed amount plus your margin, with the arithmetic on the receipt.

How do I bill AI agents?

Meter whatever the agent does: steps, tool calls, requests, or outcomes. Multiple meters can live on one subscription next to a platform fee or seats, so agent pricing composes instead of fighting the billing system.

How do you handle sub cent prices without rounding errors?

All usage math runs in exact decimal arithmetic and rounds exactly once, at the invoice line total. A $0.0000025 per token rate stays exact across two hundred million tokens, and the trace shows the unrounded intermediate values.

Will my customers trust usage based invoices?

That is the point of the receipt. Every charge carries a calculation trace showing the meter reading, the rate, and the math, so a customer can verify an invoice instead of disputing it.