generated: '2026-09-04'
method: searched
source: >-
  https://getanyapi.com/docs, https://getanyapi.com/docs/sdks,
  https://getanyapi.com/docs/mcp-server and openapi/anyapi-gateway-openapi.json
limit_count: 0
limit_count_note: >-
  AnyAPI publishes NO rate limit of its own - no requests-per-second, per-minute or per-day
  figure appears anywhere in the docs, the OpenAPI or the site, and no RateLimit-* or
  X-RateLimit-* response headers are declared or observed. This is an honest zero, and it is
  structural rather than an oversight: throughput here is governed economically, by the wallet
  balance, per-key spend limits and the per-request max_cost_usd ceiling, instead of by a
  request quota. An agent's budget is the limiter.
limits: []
exhaustion_semantics:
  economic:
    - {status: 402, code: insufficient_balance, meaning: wallet balance too low, billed: false}
    - {status: 402, code: key_cap_exceeded, meaning: the API key reached its configured spend cap, billed: false}
    - {status: 402, code: grant_cap_exceeded, meaning: the OAuth authorization reached its owner-approved spending limit, billed: false}
    - {status: 402, code: trial_cap_reached, meaning: the autonomous trial reached its cap, billed: false, remediation: response carries an upgrade link and a live RFC 8628 device authorization}
  upstream:
    - status: 429
      code: provider_rate_limited
      meaning: >-
        The UPSTREAM data source rate-limited the request - NOT an AnyAPI caller quota. Inspect
        the `payment` object when present before deciding whether to retry.
      billed: false
response_headers:
  rate_limit_headers: []
  rate_limit_headers_note: None. Neither the IETF RateLimit-* draft headers nor X-RateLimit-* are declared in the spec or returned.
  retry_after:
    present: true
    scope: >-
      Retry-After is returned on 202 durable-request responses and on 409
      idempotency_in_progress, i.e. as a work-in-progress backoff signal rather than a
      quota-exhaustion signal. The durable-request body carries the same value as
      retryAfterSeconds.
client_behavior:
  sdks: >-
    The first-party TypeScript and Python SDKs map 429 to RateLimitedError and retry it
    automatically (429 and 5xx are the only statuses they retry).
  docs: https://getanyapi.com/docs/sdks
concurrency:
  documented_limit: null
  note: >-
    No concurrency cap is published. Concurrent calls carrying the same Idempotency-Key, SKU
    and input are collapsed onto one active Request rather than rejected.