generated: '2026-08-13'
method: derived
source: "@tofuhq/enrich@0.7.3 dist/index.js (published npm bundle) + live response
  headers from https://api.enrich.tofuhq.com/v1/schema/company"
api: https://api.enrich.tofuhq.com
limit_count: 0
rate_limits: []
response_headers:
- header: x-credits-charged
  description: Credits consumed by this request. Read by the CLI off every response.
  observed: read by the client bundle; not present on free/unauthenticated endpoints
- header: x-credits-remaining
  description: Remaining credit balance after this request.
  observed: read by the client bundle; not present on free/unauthenticated endpoints
- header: x-enrich-notice
  description: Out-of-band notice string surfaced to the caller.
  observed: read by the client bundle
- header: Retry-After
  description: NOT used. The retry hint is carried in the JSON body instead.
  present: false
body_signals:
- field: error.retry_after
  description: >-
    Seconds to wait before retrying, carried inside the error envelope rather than
    in an HTTP header. This is the only backoff signal the API emits.
  applies_to_codes: [rate_limited, provider_rate_limited, provider_unavailable]
exhaustion:
  rate_limit_codes: [rate_limited, provider_rate_limited, provider_unavailable]
  rate_limit_status: 429 (503 for provider_unavailable)
  cli_exit_code: 3
  quota_codes: [budget_exhausted, free_tier_exhausted]
  quota_cli_exit_code: 4
  note: >-
    Tofu separates rate limiting from quota exhaustion. Quota is metered in
    CREDITS, not requests — see plans/tofu-plans-pricing.yml for per-operation
    credit costs. Credits are only charged for records actually returned.
batch_limits:
- operation: company get
  limit: 25 identifiers per request
- operation: person get
  limit: 25 identifiers per request
- operation: company search / person search
  limit: paginated; 100 rows per credit-charged page, `--all` auto-paginates
notes: >-
  No numeric rate limit is published anywhere — not on enrich.tofuhq.com, not in
  the shipped GUIDE.md, and not in the published Agent Skill. limit_count is 0 as
  an honest zero. What Tofu DOES ship is the runtime signal: a distinct
  `rate_limited` error code, a `retry_after` value in the error body, and
  per-request credit accounting on `x-credits-charged` / `x-credits-remaining`
  headers. The unauthenticated /v1/schema/company response carried no
  X-RateLimit-* or RateLimit-* headers of any kind.