generated: '2026-08-30'
method: searched
source: https://closedloop.sh/docs/api-reference/conventions
source_status: 200
secondary_source: https://closedloop.sh/docs/account/api-keys
summary: >-
  ClosedLoop AI publishes NO numeric per-key quota for the /v1 REST API and returns NO rate-limit
  response headers. It states plainly that there is no general fixed per-key quota, and that the
  only throttling is an implementation-defined protective safeguard on SEMANTIC search, enforced
  independently per API replica rather than as a fleet-wide allowance. Clients are told to handle
  429 and honour Retry-After. That is a documented posture, not an undocumented gap -- but it means
  an agent has no runtime budget signal to read, only a 429 to react to.
limits:
- scope: per-key, general
  window: null
  limit: null
  burst: null
  documented: true
  detail: >-
    "The public API has no general fixed per-key quota." Also restated on the API Keys page: "The
    public API currently has no fixed published per-key quota."
- scope: semantic search (GET /themes and GET /features, shared)
  window: null
  limit: null
  burst: null
  documented: true
  enforcement: per-replica
  shared_budget: true
  detail: >-
    Semantic search uses implementation-defined protective rate and concurrency safeguards. /themes
    and /features share ONE semantic-search budget and are not separately allowanced, so semantic
    traffic on either endpoint can produce a 429 on the other. Safeguards are enforced independently
    by API replica and are explicitly NOT a fleet-wide numeric quota. No number is published.
limit_count: 0
limit_count_note: >-
  Zero PUBLISHED NUMERIC limits. Two documented throttling scopes exist but the provider publishes
  no value for either, by design.
response_headers:
  documented: []
  documented_note: >-
    No X-RateLimit-*, RateLimit-* or equivalent header is documented anywhere in the API reference,
    and none appears in the OpenAPI response definitions for the 429. The ONLY documented runtime
    signal is Retry-After on 429 and on the 503 backoff codes.
  retry_after:
    documented: true
    applies_to:
    - status: 429
      code: RATE_LIMIT_EXCEEDED
      guidance: honour Retry-After
    - status: 503
      code: API_KEY_AUTH_UNAVAILABLE
      guidance: honour Retry-After -- wait 10 seconds, then exponential backoff if failures continue
    - status: 503
      code: EXPORT_UNAVAILABLE
      guidance: retry shortly
    - status: 503
      code: SEMANTIC_SEARCH_UNAVAILABLE
      guidance: retry, or opt in to lexical fallback with allow_fallback=true
exhaustion:
  status: 429
  code: RATE_LIMIT_EXCEEDED
  envelope: '{"error": "...", "code": "RATE_LIMIT_EXCEEDED", "hint": "..."}'
  declared_on_operations:
  - listThemes
  declared_on_operations_note: >-
    Only GET /themes declares a 429 in the published OpenAPI. GET /features shares the same semantic
    budget per the Conventions page but its spec entry omits the 429 -- a spec/docs disagreement
    captured in overlays/closedloop-public-api-overlay.yaml.
pagination_ceiling:
  max_page_size: 200
  default_page_size: 50
  detail: limit defaults to 50 and is capped at 200 on offset-paginated list endpoints.
quota:
  model: credits
  detail: >-
    Volume is metered as CREDITS on ingestion (1 credit = 20 characters processed), not as API
    request counts. Credit consumption is readable through GET /usage. See
    plans/closedloop-plans-pricing.yml.
  usage_endpoint: GET /usage