Usage, billing and quotas · Docs

Section 12

12. Usage, billing and quotas

Every route below resolves your tenant from the credential — no ids in URLs.

Endpoint	Answers
`GET /v1/usage`	spend over 24h / 7d / 30d / all-time, plus per-kind breakdown
`GET /v1/billing/costs`	per-request ledger rows, newest first; paged by `X-Nozzle-Next-Cursor`, filterable by time, kind, model, provider, key, project
`GET /v1/billing/costs/{request_id}`	the ledger rows one call produced
`GET /v1/analytics/usage`	calls, errors, tokens, cost and p50/p95 latency grouped by time and/or model, provider, key, project or kind
`GET /v1/billing/balance`	prepaid credit balance
`GET /v1/billing/credits`	credit ledger, paginated
`GET /v1/billing/entitlement`	your plan and any overrides
`GET /v1/quotas`	one slot per quota kind: usage and ceiling in one stated unit (see below)
`GET /v1/analytics/cost`	per-day cost series
`GET /v1/analytics/instances`	per-instance lifetime and spend
`GET /v1/analytics/margin`	inference versus compute spend

Each has an addressed twin at /v1/tenants/{tenant_id}/… for principals that span tenants. They return identical data; the caller-scoped form exists because the id could only ever have been your own.

Quotas

GET /v1/quotas always returns exactly one slot per quota kind, read through the same code path that refuses a request, so what you read is what you are enforced on. Within a slot, value and ceiling are in the same unit.

json

{"tenant_id":"…","plan_id":"enterprise","slots":[
  {"kind":"monthly_credits","period_key":"2026-09","value":155640,"ceiling":null,
   "unit":"micro_cents","measure":"spend","unlimited":true,"source":"api.cost_ledger","enforced":true},
  {"kind":"max_concurrent_instances","period_key":"","value":0,"ceiling":null,
   "unit":"instances","measure":"gauge","unlimited":true,"source":"compute.instances","enforced":true},
  {"kind":"per_minute","period_key":"2026-09-21T14:05","value":12,"ceiling":600,
   "unit":"requests","measure":"counter","unlimited":false,"source":"redis:quota_bucket","enforced":true},
  {"kind":"max_model_size_b","period_key":"","value":0,"ceiling":9999,
   "unit":"params_b","measure":"ceiling_only","unlimited":false,"source":"api.plans","enforced":true}
]}

monthly_credits — this calendar month's spend summed from your cost ledger, against your plan allowance, both in micro-cents (1¢ = 1,000,000 µ¢).
max_concurrent_instances — a live gauge of instances holding hardware.
per_minute / per_day — your plan's request rate across every inference call from every key in the tenant, as rolling windows; value is how much of the window is in use now. Refused with 429 rate_limit.exceeded and details.remaining_ms. The same ceilings appear in the limits block of GET /v1/models.
max_model_size_b — bounds a launch request; value is always 0 and is not a measurement (measure: "ceiling_only").
ceiling: null appears only with unlimited: true (an enterprise override).

Plans

Plan	Monthly credits	Max concurrent instances	Requests / minute	Requests / day
free	0¢	1	30	1,000
hobby	2,000¢	1	60	5,000
pro	20,000¢	5	600	50,000
team	100,000¢	20	6,000	500,000
enterprise	1,000,000¢	9,999	unlimited	unlimited

Exceeding the monthly allowance returns 429 budget.exceeded. The prepaid balance is deliberately not enforced per call.