AI billing that reconciles to the source
Bill AI on true token economics — including cache tokens — reconcile our figure against Anthropic’s own Admin-API cost report nightly with a drift %, and meter usage to Stripe.
If you resell or internally charge for AI, estimates drift from the provider’s invoice. Cache tokens are billed separately, and without reconciliation your margin is a guess.
Built to do the work, not just show it
True token economics
Price input, output, and cache-creation/read tokens the way the provider does.
Nightly reconciliation
Compare our aggregate against Anthropic’s Admin-API cost report per UTC day, with a drift %.
Stripe usage metering
Meter tenant AI usage to Stripe for billing.
Graceful degradation
No admin key or meter configured? It simply no-ops.
From signal to result
Capture
Every AI call logs priced token usage, cache included.
Reconcile
A nightly job books our cost vs Anthropic’s and computes drift.
Meter
Usage flows to Stripe for invoicing.
- Cache-token pricing fixed a real cost under-report.
- A model-catalogue guardrail prevents billing phantom/unpriced model IDs.
- Reconciliation is org-level against the provider’s own figure.
Questions
How accurate is the AI cost?
It’s reconciled nightly against Anthropic’s Admin-API cost report, with the drift surfaced.
Do you handle cache tokens?
Yes — they’re captured and priced separately, which is where naive meters under-report.