generated: '2026-08-13'
method: derived
source: >-
  openapi/_original/gleefulai-visibility-openapi.json, llms/gleefulai-llms.txt,
  plans/gleefulai-pricing.json
limit_count: 0
limits: []
headers: []
status_on_exhaustion: null
note: >-
  No rate limits are published anywhere in the harvested surface. Zero matches for
  RateLimit-*, X-RateLimit-*, Retry-After or 429 across the OpenAPI (36 operations), the
  llms.txt and the machine-readable pricing document, and no 429 response is declared on
  any operation. This is an honest zero derived from the provider's own contract, not an
  unchecked gap. The docs host could not be re-read on this pass because the domain is
  now parked for sale (see lifecycle/gleefulai-lifecycle.yml), so `method` stays
  `derived` rather than `searched`.
rationing_model: >-
  Consumption was rationed by price rather than by quota: every priced call answered 402
  with an x402 challenge and cost $0.01-$0.55 in USDC on Base, so cost — not a request
  ceiling — was the throttle. The free discovery and preview endpoints
  (/api/health, /api/status, /api/catalog, /api/pricing, /api/capabilities,
  /api/examples/{endpoint}, /api/preview/audit, /api/preview/compare) carried no
  documented limit at all.
gap: >-
  An agent had no runtime signal for throttling: no limit headers, no 429 contract, and
  no documented Retry-After. Combined with the absence of idempotency
  (conventions/gleefulai-conventions.yml), a retry after a timeout risked a second
  charge with no way to detect back-pressure.