Skip to main content
All of these need usage:read. Days are UTC.

Metered usage

GET /v1/usage?from=2026-10-01&to=2026-10-31&group_by=day,kind gives daily totals of every usage event: jobs submitted and finished by status, documents parsed, pages read (text or image), model tokens and cost, emails received, ignored, merged and routed, webhook deliveries and stored bytes. Totals are rolled up every 5 minutes. from and to default to the last 30 days; group_by takes day, kind or both.

Model cost in detail

GET /v1/usage/luna?days=30 gives calls, paid and replayed calls, prompt, cached, completion and reasoning tokens, and the provider-reported cost, grouped by day, task and model. Add job_id= for one job. Each job’s result also carries its own usage.
Replayed calls are answered from the cache: a document or request seen before costs nothing the second time.

Reading quality

GET /v1/quality?from=&to= reports, per bank: statements read, how many verified, image reads, scans refused as unreadable, labelling and second-read verification, the frequent review reasons, and the model cost per verified statement. Use it to see which banks’ statements need better copies from your ISOs.