generated: '2026-09-13'
method: searched
source: https://docs.squared.ai/api-reference/introduction
docs: https://docs.squared.ai/api-reference/introduction
limit_count: 2
summary: >-
  One platform-wide REST limit is published in the API reference, and one endpoint-class limit is
  published in the September 2026 release notes. Neither is accompanied by response headers, so an
  agent has no runtime signal - it can only observe the 429.
rate_limits:
- name: REST API default
  scope: api-wide
  window: 1 minute
  limit: 100
  unit: requests
  burst: null
  status_on_exhaustion: 429
  source: https://docs.squared.ai/api-reference/introduction
  quote: >-
    "The AI Squared API is rate limited to 100 requests per minute. If you exceed this limit, you
    will receive a 429 Too Many Requests response."
  note: >-
    The docs do not say whether the 100/minute is counted per API key, per user, per workspace or
    per IP.
- name: Agentic Coding Git session events
  scope: per-ip
  window: 1 minute
  limit: 60
  unit: requests
  status_on_exhaustion: 429
  source: https://docs.squared.ai/release-notes/2026/September-2026.md
  quote: '"Requests are rate limited to 60 per minute per IP."'
  note: >-
    Applies to the Git session-events endpoint added in September 2026 for Agentic Coding
    sessions. That endpoint is not in the published OpenAPI. The same release note records that
    Git and webhook endpoints generally are now rate limited, without stating those numbers.
response_headers:
  published: false
  headers: []
  note: >-
    No X-RateLimit-*, RateLimit-* or Retry-After header is documented anywhere in the docs, and no
    operation in the OpenAPI declares a 429 response or any rate-limit header. The limits above
    are prose-only. Verified against the published contract on 2026-09-13; not verified against a
    live response, because every operation requires a JWT and this pass is anonymous.
related:
  connector_level: >-
    Separately from the platform API limit, a destination CATALOG carries its own outbound
    throttle - catalog.request_rate_limit, catalog.request_rate_limit_unit and
    catalog.request_rate_concurrency (example values 600 per minute, concurrency 10) in the
    POST/PUT /api/v1/catalogs request body. That is AI Squared rate-limiting itself when it writes
    to a customer's destination, not a limit on callers of this API.
  source: openapi/ai-squared-openapi.yml