generated: '2026-08-13'
method: searched
source: >-
  https://api.involve.asia/llms.txt and https://api.involve.asia/docs/openapi.yaml
  (both HTTP 200, first-party) plus https://api.involve.asia/skill.md — the
  provider's own machine-readable docs artifacts, which are where Involve Asia
  publishes its throttle numbers.
docs: https://api.involve.asia/docs/
description: >-
  Published rate limits for the Involve Asia Publisher API. Involve Asia
  documents the numbers but returns NO rate-limit response headers — an agent
  gets the 429 status and nothing else, so client-side accounting is the only
  way to stay under the throttle.
limit_count: 2
limits:
  - name: account-throttle
    scope: per-account
    applies_to: all endpoints
    window: 1 minute
    limit: 60
    unit: requests
    burst: null
    status_on_exhaustion: 429
    note: >-
      60 requests / minute / account across every endpoint, including
      /authenticate. Documented in the OpenAPI info.description
      ("Throttle: 60 requests / minute per account") and in llms.txt.
  - name: deeplink-cap
    scope: per-account
    applies_to: POST /deeplink/generate
    window: rolling 30 days
    limit: 1000
    unit: unique tracking links
    burst: null
    status_on_exhaustion: 429
    note: >-
      1,000 unique tracking links per rolling 30-day window per account. This is
      a quota, not a throttle — it does not reset on a fixed boundary, and
      repeated /deeplink/generate calls with the same inputs may mint additional
      links against the cap because the operation has no idempotency key.
response_headers:
  rate_limit_headers: none
  documented:
    - name: request_id
      purpose: >-
        Per-response correlation id. Not a rate-limit signal — Involve Asia
        support asks for it (plus the first 6 chars of your API key) when you
        file a ticket.
  note: >-
    No X-RateLimit-*, RateLimit-* (RFC 9331 draft) or Retry-After headers are
    documented or observed. A client cannot read remaining quota or reset time
    from a response; it must count its own calls.
exhaustion:
  status: 429
  body: >-
    Standard {status, message, data} envelope; no problem+json. See
    errors/involve-asia-problem-types.yml.
  retry_guidance: >-
    Provider-published backoff ladder — 250 ms, then 500 ms, then 1000 ms.
    Verbatim from https://api.involve.asia/skill.md and llms.txt.
related:
  conventions: conventions/involve-asia-conventions.yml
  errors: errors/involve-asia-problem-types.yml
  authentication: authentication/involve-asia-authentication.yml
x-evidence:
  - url: https://api.involve.asia/llms.txt
    status: 200
    fetched: '2026-08-13'
  - url: https://api.involve.asia/docs/openapi.yaml
    status: 200
    fetched: '2026-08-13'
  - url: https://api.involve.asia/skill.md
    status: 200
    fetched: '2026-08-13'