generated: '2026-09-03'
method: probed
source: https://fluentedi.com/.well-known/ai
docs: https://fluentedi.com/llms.txt
limit_count: 0
note: >-
  An honest zero with a reason. The provider does not merely omit rate limits - it declares their
  absence twice in machine-readable form: /.well-known/ai carries "rate_limit": "none" and llms.txt
  states "Free and unauthenticated, with no rate limit and no paid tier". A live response was
  inspected and carries no X-RateLimit-*, RateLimit-* or Retry-After header, which is consistent.
  The practical caveat for an agent: because no limit is published and no headers are returned,
  there is also no runtime signal if the edge does throttle - the origin is behind Cloudflare, whose
  own DDoS protection is not part of the provider's declaration. An agent should treat a 429 as
  possible and unannounced rather than impossible.
declared_limits: []
response_headers:
  ratelimit_headers_present: false
  observed_on:
    url: https://fluentedi.com/v1/time/now?timezone=UTC
    status: 200
    headers_returned:
    - access-control-allow-origin
    - cache-control
    - link
    - strict-transport-security
    - x-content-type-options
    - x-docs
    - x-frame-options
    - x-service
    - cf-ray
    - server
  retry_after: not observed
  status_on_exhaustion: not published
structural_limits:
- scope: per-request
  surface: POST /v1/batch
  limit: 20
  unit: tool calls per batch
  enforcement: 'HTTP 400, error code too_many_calls, message "A batch holds at most 20 calls; this one has N. Split it."'
  verified: probed
  source: observed live on 2026-09-03 by posting 25 calls
- scope: per-request
  surface: doc.extract
  limit: 10
  unit: MB per document
  enforcement: documented in the tool description
  verified: searched
  source: https://fluentedi.com/llms-full.txt