Skip to content

Per-token billing for AI products

You pay the inference cost of every token your AI product serves. At a flat price, one heavy customer can put you at a loss. Payrudder meters tokens, images or minutes per customer, warns you at the thresholds you set and turns the totals into invoices.

The hard part

  • 01

    Inference cost grows with every request while the flat plan stays the same, and the gap comes out of your margin

  • 02

    Clients retry, and one duplicated batch means a customer is billed twice

  • 03

    Customers want to see their usage before the invoice arrives

How Payrudder handles it

  1. Meter every unit you sell

    Define meters such as input tokens, output tokens or images and send events as they happen. Quantities are stored as exact decimals, not floats.

    Meter every unit you sell
  2. One alert per threshold

    Entitlement checks answer in milliseconds and each threshold fires once. Warn the customer before the invoice, or stop usage in your app.

    One alert per threshold
  3. Mix a base fee with usage

    Put usage, graduated tiers or credit packs on top of a base fee. If your prices change, you can re-rate a closed month.

    Mix a base fee with usage

Questions

Can I bill input and output tokens at different prices?

Yes. Make a meter for each, or filter one meter by a property such as model, and price them separately in the package.

Does Payrudder stop requests when a customer runs out?

No, blocking is your app's call. The entitlement check tells your app whether the customer is within limits. Incoming usage events are kept even over the limit.

Start billing by usage

The free plan covers your first live payments. Hook up one meter, close a month in test mode, then go live.