generated: '2026-08-12'
method: searched
source: >-
  https://developer.flora.ai/platform/errors, https://developer.flora.ai/mcp/tools,
  https://developer.flora.ai/mcp/troubleshooting,
  https://developer.flora.ai/recipes/batch-from-csv and
  https://developer.flora.ai/recipes/upload-an-asset — every page on FLORA's
  developer site that mentions rate limiting was read. None states a number.
provider: FLORA
providerId: flora-fauna
description: >-
  FLORA enforces rate limits and tells you how to react to them, but never says
  what they are. The 429 response, the error code, the header and the scope are all
  documented; the limit, the window and the burst are not published anywhere. The
  guidance the provider gives instead is empirical — "drop to 2 concurrent",
  "bound uploads at 4 at a time" — which is the shape of advice you give when the
  number is not public.
limit_count: 0
published_limits: false
limits: []
scope: workspace
scope_note: >-
  Limits apply at the workspace level and are shared between the REST API and the
  MCP server — the docs state MCP "rate limits apply at the workspace level and
  match the REST API". An agent session and a backend batch job therefore contend
  for the same budget, which is a real operational hazard the docs do not price.
responseCodes:
  throttled: 429
  error_code: rate_limited
headers:
  retryAfter: retry-after
  retryAfter_reliability: >-
    Documented as "when present". FLORA's own published retry sample falls back to
    5 seconds when the header is absent, so a client cannot depend on it.
  requestId: request-id
  rateLimitRemaining: null
  rateLimitLimit: null
  rateLimitReset: null
  standard_ratelimit_headers: false
  note: >-
    No RateLimit-* or X-RateLimit-* headers are documented. A client learns it is
    near a limit only by being refused — there is no forward-looking runtime signal
    to pace against, which is the single most useful thing an agent needs here.
provider_guidance:
  - context: batch from CSV
    advice: 'If you see 429 rate_limited, drop concurrency to 2 and let exponential backoff handle the rest.'
    source: https://developer.flora.ai/recipes/batch-from-csv
  - context: bulk asset upload
    advice: 'For 50+ files, run uploads in parallel with bounded concurrency (e.g. 4 at a time) to avoid rate limits.'
    source: https://developer.flora.ai/recipes/upload-an-asset
  - context: MCP interactive use
    advice: '429 is "rare in interactive use, more common in batch jobs" — reduce parallelism.'
    source: https://developer.flora.ai/mcp/troubleshooting
  - context: retry policy
    advice: 'Retry 429, 500 and 503 with exponential backoff capped at 30s; never retry 400/401/403/404/409/422.'
    source: https://developer.flora.ai/platform/errors
plan_linked_limits:
  published: false
  note: >-
    The pricing pages tie plans to a monthly USD generation budget, not to request
    rates. The effective throughput ceiling most integrators will hit first is
    financial (402 insufficient_credits), not a 429.
adjacent_quotas:
  - {name: seats, limit: 8, scope: self-serve workspace, source: 'https://docs.flora.ai/plans-and-billing/pricing'}
  - {name: active API keys, limit: 1, scope: workspace, source: 'https://developer.flora.ai/platform/authentication'}
  - {name: MCP execute HTTP request timeout, limit: 30s, scope: per request inside the sandbox, source: 'https://developer.flora.ai/mcp/tools'}
  - {name: MCP execute total run time, limit: ~5 minutes, scope: per execute call, source: 'https://developer.flora.ai/mcp/tools'}
  - {name: webhook delivery timeout, limit: 10s, scope: per delivery attempt, source: 'https://developer.flora.ai/platform/webhooks'}
observed: false
observed_note: >-
  Not measured against a live endpoint. Unauthenticated calls to
  https://app.flora.ai/api/v1/techniques and /models return 401 before any rate
  limit applies, so no headers could be read anonymously and no probing was
  attempted against a credentialed account.