generated: '2026-08-14'
method: searched
source: https://docs.lang.ai/#rate-limit
checked: '2026-08-14'
summary: >-
  Lang.ai documents that rate limiting EXISTS but publishes not one number and not one
  header. The entire published contract is a single sentence: "To avoid abuse and ensure
  that the system remains online for every user, every API request will need to provide
  valid credentials and the requests can be rate limited if they exceed the limitations."
  The error table then lists 429 "Too many requests / Rate limited". There is no window,
  no quota, no burst, no per-key vs per-account scope, no RateLimit-*/X-RateLimit-*
  response header and no Retry-After. An agent cannot budget against this API; it can only
  react to a 429 it has already earned.
limit_count: 0
limit_count_note: >-
  An honest zero. Rate limiting is documented as enforced but no limit is quantified
  anywhere in the reference, the help center, or any Lang.ai surface probed.
enforced: true
scopes: []
headers: []
headers_note: >-
  No rate-limit response headers are documented. The live surface could not be observed:
  the base URL is the per-tenant template https://{company}.lang.ai/api/v1 and there is no
  public tenant to probe anonymously, so no 429 could be induced or inspected. Absence
  here means "not published and not observable", not "not emitted".
status_on_exhaustion: 429
status_source: 'https://docs.lang.ai/#errors — "429 | Too many requests / Rate limited"'
retry_after: null
retry_guidance: >-
  With no window, no remaining counter and no Retry-After, the only safe client behavior
  is exponential backoff with full jitter on 429, plus a client-side concurrency cap.
  Retries of saveDocument are safe when a stable client-supplied document `id` is sent
  (see conventions/langai-conventions.yml); createProject has no idempotency contract and
  must be reconciled against listProjects before a blind retry.
quotas: []
concurrency: null
evidence:
  - url: https://docs.lang.ai/
    http_status: 200
    section: Rate Limit
    quote: >-
      "To avoid abuse and ensure that the system remains online for every user, every API
      request will need to provide valid credentials and the requests can be rate limited
      if they exceed the limitations."
  - url: https://docs.lang.ai/
    http_status: 200
    section: Errors
    quote: '429 — Too many requests / Rate limited'
  - url: https://company.lang.ai/api/v1/projects
    http_status: null
    note: >-
      Not probed. `company` is a documentation placeholder for the customer's own tenant
      subdomain, not a live host — see security/langai-domain-security.yml, where the TLS
      probe records a hostname-mismatch certificate error for company.lang.ai.
related:
  conventions: conventions/langai-conventions.yml
  errors: errors/langai-problem-types.yml
  plans: plans/langai-plans-pricing.yml