generated: '2026-09-18'
method: searched
source: https://docs.neuralverge.ai/errors-and-rate-limits
scope_model: per-organization
scope_note: >-
  Limits apply per organization, shared across every API key and user in the
  org (not tracked per key). Higher org-wide limits can be configured on request.
enforcement: queue
enforcement_note: >-
  Instead of rejecting on the instant a limit is hit, requests are queued (oldest
  first, per endpoint) and started as capacity frees. A request only fails with
  429 if still waiting after 60 seconds in the queue; a 429 therefore means the
  request never ran and is always safe to retry.
exhaustion_status: 429
response_headers: []
response_headers_note: No rate-limit response headers (X-RateLimit-*, RateLimit-*, Retry-After) are documented.
limit_count: 5
limits:
  - scope: per-organization
    dimension: concurrent-requests
    limit: 50
    window: null
    applies_to: all endpoints
  - scope: per-organization
    dimension: requests-per-second
    limit: 10
    window: 1s
    applies_to: all endpoints
  - scope: per-endpoint-per-organization
    dimension: requests-per-second
    limit: 20
    window: 1s
    applies_to: default per endpoint
  - scope: per-endpoint-per-organization
    dimension: requests-per-second
    limit: 10
    window: 1s
    applies_to: run-email-enrichment
  - scope: per-endpoint-per-organization
    dimension: requests-per-second
    limit: 10
    window: 1s
    applies_to: [run-phone-enrichment, run-phone-enrichment-us]
handling_429:
  - Back off and retry (the queue timeout means the request never ran, so retry is safe)
  - Use exponential backoff, not a tight retry loop
  - Spread requests over time instead of bursting if you regularly hit the queue timeout