How do I charge AI app users by token usage?

Tokens or cents?

Use cents if you support multiple models with different costs. Use tokens if you only call one model or charge a flat rate per token. Both work; cents is more flexible.

Reserve an upper bound, then refund

Use a tokenizer (tiktoken for OpenAI) to estimate prompt tokens, then reserve prompt + max_tokens. After the call, commit and refund the unused portion.

Pre-paid credits

On Stripe one-time checkout, bump the user's custom limit by the purchased amount. Their counter keeps ticking from the new ceiling.

await vevee.upsertSubscription({
  userId,
  planId: "plan_paygo",
  customLimits: { tokens: { quota: currentQuota + addedTokens } },
});

Live balance display

Use a pk_live_ public key in the browser. vevee.usage(userId) returns the user's remaining quota. Safe to expose - only reads the calling user's own counters.