generated: '2026-09-11'
method: searched
source: https://www.socialfetch.dev/docs/credits.mdx and docs/errors.mdx (read 2026-09-11)
model: credit-metered (no request-per-second quota on paid routes)
limit_count: 2
limits:
  - scope: per-key (free routes only)
    applies_to: [GET /v1/whoami, GET /v1/balance]
    window: unspecified
    limit: unspecified
    status_on_exhaustion: 429
    code: temporarily_unavailable
    detail: Free auth/balance routes enforce a per-key rate limit and return 429 when exceeded.
  - scope: concurrency (paid metered routes)
    window: instantaneous
    limit: "~500 concurrent requests (recommended ceiling, not a hard quota)"
    status_on_exhaustion: 503
    code: temporarily_unavailable
    detail: >-
      Paid metered /v1/* routes have no request quota — credits are the only
      limit. Under extreme load the API returns 503 with Retry-After (not
      charged). Stay below ~500 concurrent requests.
response_headers:
  ratelimit_headers: []
  retry_after:
    trigger_status: 503
    detail: Present on 503 temporarily_unavailable; honor it before retrying.
  note: >-
    No X-RateLimit-* / RateLimit-* headers are documented or returned; balance is
    polled via GET /v1/balance (data.balance, data.billingAlert).
exhaustion_credits:
  status: 402
  code: insufficient_credits
  detail: When the credit balance is empty, metered routes return 402 insufficient_credits.
notes: >-
  The runtime signal for capacity is credits (meta.creditsCharged, GET
  /v1/balance) rather than rate-limit headers; 429 is scoped to the free routes,
  503+Retry-After signals transient overload.