specification: API Commons Rate Limits
specificationVersion: '0.1'
schema: https://raw.githubusercontent.com/api-evangelist/interface-research/main/schema/api-commons.yml#/$defs/RateLimits
provider: Caterpillar
providerId: caterpillar
created: '2026-05-04'
generated: '2026-09-05'
method: searched
modified: '2026-09-05'
reconciled: true
source: https://digital.cat.com/knowledge-hub/articles/iso-15143-3-aemp-20-api-developer-guide
tags:
  - Telematics
  - Heavy Equipment
  - Rate Limiting
  - ISO 15143-3
description: >-
  Caterpillar publishes real numeric limits for the ISO 15143-3 (AEMP 2.0) API in
  its developer guide. Access is scoped to the OAuth client id issued on
  subscription approval: a daily quota of 50,000 requests per client id shared
  across all endpoints, resetting at 12:00am GMT, plus a per-second throttle. The
  guide gives a general throttle of 10 calls per second per client id with a
  recommendation to stay at 7-8, and a higher 100 calls per second throttle for
  the Fleet Snapshot, Equipment Snapshot and Fault Codes Timeseries endpoints.
  Exhausting the quota returns HTTP 429 with the gateway's
  {"code":"429.002","description":"Quota rate limit exceeded."} envelope.
notes: >-
  The numbers are documented but the runtime signal is not. No X-RateLimit-*,
  RateLimit-* or Retry-After response header is documented, and none was present
  on anonymous responses observed from https://services.cat.com on 2026-09-05, so
  a client cannot read its remaining budget at runtime and has to track the quota
  itself. Limits for the VisionLink API family (assetSummary, faultsHistory,
  assetsOperation, assetFenceHistory, utilization) are not publicly published.
  This artifact replaces the 2026-05-04 bulk-sweep placeholder, which recorded
  "limits not published" — they are published; the earlier pass was stopped by the
  Cloudflare bot block on digital.cat.com and did not reach the developer guide.
sources:
  - https://digital.cat.com/knowledge-hub/articles/iso-15143-3-aemp-20-api-developer-guide
  - https://digital.cat.com/api-catalog-overview
responseCodes:
  throttled: 429
  throttledCode: '429.002'
  throttledBody: '{"code":"429.002","description":"Quota rate limit exceeded."}'
  serviceUnavailable: 503
responseHeaders:
  rateLimitHeaders: []
  retryAfter: false
  note: No rate-limit response headers are documented or observed.
limits:
  - name: Daily request quota
    api: caterpillar:iso-15143-3-aemp-api
    scope: per-client-id
    metric: requests
    limit: 50000
    window: day
    reset: 00:00 GMT
    shared_across_endpoints: true
    method: searched
    source: https://digital.cat.com/knowledge-hub/articles/iso-15143-3-aemp-20-api-developer-guide
  - name: General throttle
    api: caterpillar:iso-15143-3-aemp-api
    scope: per-client-id
    metric: requests
    limit: 10
    window: second
    recommended_max: 8
    method: searched
    source: https://digital.cat.com/knowledge-hub/articles/iso-15143-3-aemp-20-api-developer-guide
  - name: Snapshot / fault-code timeseries throttle
    api: caterpillar:iso-15143-3-aemp-api
    scope: per-client-id
    metric: requests
    limit: 100
    window: second
    endpoints:
      - Fleet Snapshot
      - Equipment Snapshot
      - Fault Codes Timeseries
    method: searched
    source: https://digital.cat.com/knowledge-hub/articles/iso-15143-3-aemp-20-api-developer-guide
  - name: VisionLink API limits
    api: caterpillar:visionlink-apis
    scope: per-client-id
    metric: unknown
    limit: null
    method: none
    note: Not publicly published; surfaced to the subscriber after approval.
policies:
  - name: Client-id-scoped quota
    description: >-
      Quota and throttle are evaluated per OAuth client id, not per user or per
      asset, and the daily quota is shared across every endpoint the client calls.
  - name: Backoff
    description: >-
      On 429 the API returns code 429.002 with no Retry-After. Retry with
      exponential backoff and jitter, and treat the 12:00am GMT reset as the
      earliest point a quota exhaustion clears.
  - name: Telemetry cadence
    description: >-
      VisionLink-derived data is updated on the platform's collection cadence;
      polling faster than the upstream refresh window burns quota without
      surfacing new data. Caterpillar's own recommendation to stay at 7-8 calls
      per second against a 10/sec ceiling points the same way.
limit_count: 3
maintainers:
  - FN: API Evangelist
    email: info@apievangelist.com