Inspect rate limits
Non-consuming snapshots of caller-scoped HTTP request windows: aggregate inference_generate limits (dimensioned limits excluded), inference_count, and model_list. No query parameters. Authentication binds all scopes; credentials remain bound to their organization. Null limit/remaining/reset_at means no applicable request bucket, or enforcement disabled when enabled is false. Remaining is capacity before the next request; reset_at is full replenishment as rounded-up Unix seconds; retry_after is rounded-up seconds until the next request (zero when allowed). Target exhaustion returns HTTP 200 with allowed false. Polling consumes only the existing spa_api_account_org quota, whose headers and 429 errors describe polling itself. Backend errors return 503. This is not an admission guarantee: concurrency, billing, free-tier quotas, dimensioned inference policies, and MCP are excluded.
Authorizations
Org API key as Authorization: Bearer z_sk_...

