Browse documentation
Scale and Enterprise quotas, operation costs, and monthly usage.
Concepts
Request units: every REST call and every MCP tool call costs 1.
Limited per minute and per day.Operation units ("API credits"): expensive work costs credits from a
monthly budget. Clients can never choose a cost.Concurrency: caps on in-flight generation and publishing jobs (generation/publishing).
Dimensions: limits apply per key, per account (owner, across all
keys), and per brand (monthly credits). Every applicable bucket must
pass. A rejection charges nothing.API and MCP share one allowance. Usage is reported separately, but admission adds both surfaces together. Creating another key or switching surface cannot double the
allowance.Windows are UTC calendar minute, day and month (database time). Credits
reset on the 1st of each month with no rollover.
Default tier limits
Tier | RPM / key | RPM / account | Requests/day / key | Requests/day / account | Credits/month (key, account, brand) | Concurrent jobs (key / account / brand) | Max active keys | Queue priority |
|---|---|---|---|---|---|---|---|---|
Free / Trial / Growth | 0 | 0 | 0 | 0 | 0 | 0 / 0 / 0 | 0 | — (no access) |
Scale | 300 | 600 | 50,000 | 100,000 | 500 | 10 / 10 / 5 | 20 | 5 |
Enterprise | 600 | 1,200 | 200,000 | 400,000 | 2,000 | 20 / 20 / 10 | 50 | 1 |
Additionally, publishing jobs are capped at a default of 5 in-flight operations per destination channel.
Lower queue-priority numbers run first among waiting API/MCP generation and publishing jobs. Active work is not interrupted. Priority is not an HTTP latency guarantee, and it does not apply to Website/Local job systems.
Credit costs
Operation | Credits |
|---|---|
Generate | 5 |
Generate | 2 |
Publishing preview | 1 |
Publishing operation (generic publish) | 3 |
Website | 3 per change set |
Website | 1 per resource |
Website | 1 |
Website | 0 |
Local run request | 5 |
Results refresh | 1 |
Calendar reschedule | 1 |
Calendar cancel, topic/draft/cover edits, Local status/location edits | 0 |
Reads, usage | 0 (1 request unit only) |
Admitted attempts are charged even if the downstream provider later fails. There are no automatic refunds. Idempotent replays are not charged again.
Reading your effective limits
Every applicable key, account, and brand bucket must pass before a request is admitted. Rejected requests are not charged. On rejection, 429 includes Retry-After; concurrency rejections default to 30 seconds. Contracted account overrides can change your effective limits, but cannot grant access to an ineligible plan.
Reading usage
Use GET /v1/usage or the MCP tool rankad_get_usage .