Token logs aren't invoices
You have usage in Postgres and prices in your processor, and a nightly script praying they agree. Every model swap breaks the maths.
Live - free tier open
Usage billing that
already speaks tokens.
ApiShip meters every model call, rates it against per-model prices, tracks your margin, and hands finished line items to your payment processor. Built for teams selling AI - not for a generic events warehouse.
Ledger · acct_9840 · period to date
Sample figures
Billed to · Your accountspend not selected
| Model | Dim | Tokens | Billed | Cost |
|---|---|---|---|---|
| gpt-5.5 | input | 142.8M | $4,141 | $1,428 |
| gpt-5.5 | cached | 470.4M | $3,410 | $1,176 |
| gpt-5.5 | output | 10.0M | $2,314 | $798 |
| claude-4.6 | input | 38.5M | $1,340 | $462 |
| embed-4 | input | 3.36B | $974 | $336 |
| Margin 66% | 4.02B | $12,180 | $4,200 | |
Nothing metered yet - this is the first invoice you'd be able to send.
Illustrative only · not a quote
Adjust the estimator → the ledger redraws → print it
Estimator · form 1040-T
Draw a sample ledger
Answer the questions below and the ledger redraws against your numbers. When it looks right, open an account and meter your first real call in minutes.
Open an account - free tier
No card required · 10k metered events free each month
Everyone selling AI rebuilds the same broken billing pipeline.
You can't see margin per customer
Provider cost, markup and plan credits live in three places, so nobody knows which accounts are burning your gross margin.
Generic metering doesn't speak tokens
Event pipelines make you model prompt tokens, cached tokens, tool calls and per-model rates by hand - every single time.
Schedule A · what you get
One API between your model calls and your invoices.
Line 01
Token-native meters, not generic events
Input, output, cached and reasoning tokens are first-class columns. Per-model rate cards ship with the SDK and version themselves, so a model swap is a rate-card entry rather than a migration.
6 dims
metered per call
Line 02
Cost and price on the same line
Every metered event carries provider cost next to what you charged. Margin per customer, per model and per feature becomes a single query instead of a quarterly finance project.
1 query
to margin per account
Line 03
Credits, allowances and hard caps
Prepaid credit packs, plan allowances, soft alerts and hard stops are enforced at request time through one call - before the expensive completion runs, not after the overage lands.
< 15 ms
check at request time
Line 04
Invoices you can defend in support
Every line item drills down to the request ID that produced it. Support answers "why was I charged this?" from the ledger, without paging an engineer into a billing thread.
req_id
on every line item
Line 05
Beside your payment processor, not instead of it
ApiShip computes the usage, rates it, and hands finalized line items to the processor you already use. You keep the payment stack and delete the reconciliation cron.
0 crons
left to reconcile
Line 06
Filed in an afternoon
Wrap your model client and you have metering, rating and invoicing. No data-engineering programme, no procurement cycle, no discovery call before you can read your own numbers.
< 1 day
to first invoice
Built for the team shipping the AI product
If your pricing has the word "credits", "tokens" or "per run" in it, you're who this is for.
- 01AI API and inference platforms billing per token or per second
- 02Agent and copilot products selling credit packs or run-based plans
- 03Dev-tool companies with a generous free tier and painful overages
- 04Teams outgrowing basic metered billing but too small for enterprise metering
Free tier open
Open your ledger today. First 10k events free.
Sign up, mint an API key, POST your first metered call, and draft an invoice in the same afternoon. No sales call, no card to start.