specification: API Commons Rate Limits
specificationVersion: '0.1'
schema: https://raw.githubusercontent.com/api-evangelist/interface-research/main/schema/api-commons.yml#/$defs/RateLimits
generated: '2026-09-13'
method: searched
source: https://www.withone.ai/pricing (published per-tier limits); 429 declarations read from
  openapi/integration-os-one-api-openapi.json
provider: IntegrationOS
providerId: integration-os
created: '2026-05-04'
modified: '2026-09-13'
supersedes: >-
  The 2026-05-04 bulk-sweep scaffold, which invented 10/100/1000 rpm tiers and a full X-RateLimit-*
  header set. The real published limits are below; the headers are NOT documented, and saying so is
  the point.
tags:
  - Agent Infrastructure
  - AI Agents
  - Integrations
  - MCP
  - Rate Limiting
  - Quotas
description: >-
  Published rate limits for the One API (api.withone.ai). The per-minute ceiling is the thing the
  paid tiers actually sell — every tier, Free included, carries unlimited connections and 1M monthly
  API calls, so the ladder is 100 -> 500 -> 1,000 requests per minute.
limit_count: 4
limits:
  - tier: free
    name: Free Tier Rate Limit
    scope: account
    metric: requests_per_minute
    limit: 100
    burst: null
    timeFrame: minute
    source: https://www.withone.ai/pricing
    applies:
      - One API (api.withone.ai)
  - tier: starter
    name: Starter Tier Rate Limit
    scope: account
    metric: requests_per_minute
    limit: 500
    burst: null
    timeFrame: minute
    source: https://www.withone.ai/pricing
    applies:
      - One API (api.withone.ai)
  - tier: pro
    name: Pro Tier Rate Limit
    scope: account
    metric: requests_per_minute
    limit: 1000
    burst: null
    timeFrame: minute
    source: https://www.withone.ai/pricing
    applies:
      - One API (api.withone.ai)
  - tier: enterprise
    name: Enterprise Negotiated Rate Limit
    scope: contract
    metric: requests_per_minute
    limit: -1
    burst: null
    timeFrame: minute
    source: https://www.withone.ai/pricing — "Custom"
    applies:
      - One API (api.withone.ai)
quotas:
  - tier: all-self-serve
    name: Monthly API Calls
    metric: requests_per_month
    limit: 1000000
    timeFrame: month
    note: >-
      1M included on Free, Starter and Pro alike; Enterprise is custom. Overage is billed per 1,000
      requests at $0.25 (Starter) or $0.15 (Pro); Free has no overage cell, so exceeding the included
      volume on Free is a hard stop rather than a charge.
    source: https://www.withone.ai/pricing
headers:
  documented: false
  limit: null
  remaining: null
  reset: null
  retryAfter: null
  policy: null
  note: >-
    NO rate-limit response headers are documented and none was observed on an unauthenticated
    response. This is the gap that matters for an agent: the limits are published as numbers on a
    pricing page, but nothing on the wire tells a caller how close it is to the ceiling or when the
    window resets. Neither X-RateLimit-* nor RateLimit-* nor Retry-After appears in the OpenAPI, the
    docs or the observed responses. Use exponential backoff with jitter on 429.
responseCodes:
  throttled: 429
  quotaExceeded: 402
  serviceUnavailable: 503
response_code_note: >-
  429 is declared on 39 of 248 operations and 402 (Payment Required) on 186 — plan-quota exhaustion
  surfaces as 402, not 429, on most of this API. An agent that handles only 429 will mishandle the
  commoner failure.
observed_headers:
  - url: https://api.withone.ai/openapi.json
    fetched: '2026-09-13'
    rate_limit_headers: none
  - url: https://mcp.withone.ai/mcp
    fetched: '2026-09-13'
    rate_limit_headers: none
    note: Returned x-request-id and x-one-correlation-id only.
policies:
  - name: Backoff Strategy
    description: >-
      Derived guidance, not provider text — no Retry-After is sent, so exponential backoff with
      jitter is the only safe retry strategy on 429.
    method: derived
  - name: Downstream Limits
    description: >-
      A passthrough call is also subject to the THIRD-PARTY platform's own rate limit, which One does
      not publish and cannot raise. A 429 on /v1/passthrough may originate downstream.
    method: derived
maintainers:
  - FN: Kin Lane
    email: kin@apievangelist.com