generated: '2026-08-27'
method: searched
source: https://ragflow.io/docs/http_api_reference + https://ragflow.io/#pricing-plan + live header probes of https://cloud.ragflow.io
api: ragflow-http-api
summary: >-
  RAGFlow publishes no request-rate limits and emits no rate-limit response headers. What it does
  publish is a monthly CREDIT allowance per RAGFlow Cloud tier — a usage quota, not a rate limit. Both
  facts are recorded below; the quota entries are marked kind: quota so they are not mistaken for
  runtime throttling signals.
limit_count: 3
request_rate_limits:
  documented: false
  count: 0
  note: >-
    The 238KB HTTP API reference contains no mention of rate limiting, throttling, 429 or Retry-After.
    The only occurrence of "rate-limited" in the whole document refers to a third-party You.com web
    search endpoint, not to RAGFlow itself.
response_headers:
  rate_limit_headers: []
  retry_after: false
  status_on_exhaustion: null
  probed:
  - url: https://cloud.ragflow.io/api/v1/system/healthz
    http_status: 200
    headers_returned: [server, date, content-type, content-length, vary, via]
    rate_limit_headers_present: false
  - url: https://cloud.ragflow.io/api/v1/datasets?page=1&page_size=1
    http_status: 401
    headers_returned: [server, date, content-type, content-length, vary, via]
    rate_limit_headers_present: false
  note: >-
    Neither X-RateLimit-*, RateLimit-* (RFC 9331 style) nor Retry-After appears on live responses. An
    agent has no runtime signal of remaining budget from this API at all.
limits:
- scope: per-account
  kind: quota
  plan: Free
  window: month
  limit: 500
  unit: credits
  burst: null
  source: https://ragflow.io/#pricing-plan
- scope: per-account
  kind: quota
  plan: Starter
  window: month
  limit: 5000
  unit: credits
  burst: null
  source: https://ragflow.io/#pricing-plan
- scope: per-account
  kind: quota
  plan: Pro
  window: month
  limit: 20000
  unit: credits
  burst: null
  source: https://ragflow.io/#pricing-plan
other_published_ceilings:
- name: dataset storage
  values: {Free: 0.1 GB, Starter: 5 GB, Pro: 50 GB}
  source: https://ragflow.io/#pricing-plan
- name: page_size (Retrieve chunks / list endpoints)
  value: 'top_k maximum 1024 on retrieval; MCP ragflow_list_datasets caps page_size at 1000, retrieval
    at 100'
  source: mcp/server/server.py
- name: dataset document count
  value: '>10,000 documents was a known CRUD failure mode until pagination was added to
    get_flatted_meta_by_kbs() in v0.26.2'
  source: https://ragflow.io/docs/release_notes
self_hosted:
  note: >-
    A self-hosted RAGFlow deployment has no credits and no vendor-imposed limits. Its ceilings are the
    operator's own hardware, the configured document engine and the rate limits of whichever model
    providers are wired in.
gaps:
- No request-rate limit is published for RAGFlow Cloud, so an integrator cannot size a client.
- No rate-limit or quota headers are returned, so an agent cannot back off before being cut off.
- Credit consumption per operation is undefined, so the 500/5,000/20,000 figures are not convertible
  into a number of API calls.