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: Globant
providerId: globant
generated: '2026-09-14'
method: searched
source: 'https://docs.globant.ai/en/wiki?802,Usage+Limits+API, https://docs.globant.ai/en/wiki?23,Error+Codes, https://glob.ai/docs/api.md'
created: '2026-05-04'
modified: '2026-09-14'
reconciled: true
supersedes: >-
  The 2026-05-04 bulk-sweep artifact, which said Globant publishes no public API and no rate-limit
  policy. Globant Enterprise AI does publish a usage-limit surface — it is just a customer-configured
  budget API rather than a platform throttle.
tags:
  - Rate Limiting
  - Artificial Intelligence
  - Software
description: >-
  Globant publishes NO numeric rate limits and NO rate-limit response headers on any surface. What it
  publishes instead is the inverse: a full Usage Limits API that lets a CUSTOMER set their own hard and
  soft ceilings, by cost or by request count, at organization or project scope. An agent integrating
  here cannot discover what it is allowed to do — it can only be told after the fact, through a numeric
  error code, that a ceiling somebody else configured has been hit.
limit_count: 0
limits: []
limits_note: >-
  Zero published limits. Not an omission — the Usage Limits API reference, the API reference index and
  the public API docs were all read, and none states a requests-per-second, per-minute, per-day or
  burst figure for any endpoint.
response_headers:
  published: []
  checked:
    - X-RateLimit-Limit
    - X-RateLimit-Remaining
    - X-RateLimit-Reset
    - RateLimit
    - RateLimit-Policy
    - Retry-After
  note: >-
    None of these are documented anywhere in the Globant Enterprise AI or Glob.AI documentation, and
    none was observed on the unauthenticated responses probed (https://glob.ai/api/catalog,
    https://glob.ai/api/v1/health, https://api.beta.glob.ai/api/v1/health, https://api.saia.ai/).
    RFC 9457 is not in use either, so there is no problem+json retry hint. An agent's only retry signal
    is the numeric error code in the body.
responseCodes:
  throttled: null
  throttled_note: >-
    No 429 is documented. Rate-limit and quota conditions surface as numeric codes inside the
    proprietary {"errors":{"id":n,"description":"..."}} envelope; the docs do not state which HTTP
    status accompanies them.
exhaustion_signals:
  - {code: 151, message: Provider Rate Limit, meaning: 'The upstream model provider throttled the call.'}
  - {code: 152, message: Provider Quota Exceeded, meaning: 'The upstream model provider quota is spent.'}
  - {code: 153, message: Provider Overcapacity, meaning: 'The upstream model provider is at capacity.'}
  - {code: 154, message: Quota exceeded, meaning: 'A Glob.AI OS usage limit configured for this organization or project has been reached.'}
  - {range: '3200-3206', meaning: 'Usage-limit error class.'}
  - {range: '4000-4046', meaning: 'Permission and quota error class.'}
exhaustion_signals_source: 'https://docs.globant.ai/en/wiki?23,Error+Codes'
customer_configured_limits:
  api: Usage Limits API
  docs: 'https://docs.globant.ai/en/wiki?802,Usage+Limits+API'
  description: >-
    Customers define their own consumption ceilings. Each limit carries a soft threshold (alerting) and
    a hard threshold (enforcement), a subscription cadence and a renewal mode.
  scopes:
    - {scope: organization, metrics: [Cost]}
    - {scope: project, metrics: [Requests, Cost]}
  subscription_types: [Freemium, Daily, Weekly, Monthly]
  renewal: [Renewable, NonRenewable]
  operations:
    - 'POST /usageLimits/organizations/{organization}/limits'
    - 'GET /usageLimits/organizations/{organization}/limits'
    - 'GET /usageLimits/organizations/{organization}/limits/active'
    - 'GET /usageLimits/organizations/{organization}/limits/latest'
    - 'DELETE /usageLimits/organizations/{organization}/limits/{id}'
    - 'PUT /usageLimits/organizations/{organization}/limits/{id}/hardLimit'
    - 'PUT /usageLimits/organizations/{organization}/limits/{id}/softLimit'
    - 'PUT /usageLimits/organizations/{organization}/limits/{id}/renewalStatus'
    - 'POST /usageLimits/organizations/{organization}/projects/{project}/limits'
    - 'GET /usageLimits/organizations/{organization}/projects/{project}/limits'
    - 'GET /usageLimits/organizations/{organization}/projects/{project}/limits/active'
    - 'GET /usageLimits/organizations/{organization}/projects/{project}/limits/latest'
    - 'DELETE /usageLimits/organizations/{organization}/projects/{project}/limits/{id}'
    - 'PUT /usageLimits/organizations/{organization}/projects/{project}/limits/{id}/hardLimit'
    - 'PUT /usageLimits/organizations/{organization}/projects/{project}/limits/{id}/softLimit'
    - 'PUT /usageLimits/organizations/{organization}/projects/{project}/limits/{id}/renewalStatus'
  auth: 'Organization-scoped endpoints require $GEAI_ORGANIZATION_APITOKEN.'
policies:
  - name: Customer-set ceilings
    description: Consumption ceilings are configured by the customer through the Usage Limits API rather than imposed and published by Globant.
  - name: Upstream provider limits pass through
    description: Model-provider throttling is surfaced verbatim as codes 151-153 rather than absorbed by the platform.
  - name: Contract-governed
    description: Commercial consumption terms live in the subscription agreement, not on a public page.
maintainers:
  - FN: Kin Lane
    email: kin@apievangelist.com