Trust center
What NapKey can prove — and what it cannot
A technical account of identity, key handling, request metadata, billing and the upstream boundary. Written to be checked, not skimmed.
Independent service
NapKey is an independent API gateway and prepaid billing service. It is not Anthropic, AWS or Kiro, and does not imply endorsement by those companies.
API key handling
A new key is shown once in the console and the control plane stores its hash. The data plane currently retains the bearer credential in restricted server configuration because it must authenticate calls; management responses expose only a masked identifier. Revocation is pushed to the data plane and may enter a visible retry state if propagation fails, so treat an exposed key as usable until deletion is synchronized.
Request records
The customer ledger stores request ID, model, token categories, cost, status and latency. Full prompts are not written to production request debug logs. Requests still pass through NapKey and the configured upstream provider. Responses API files become eligible for deletion after 30 days; an hourly cleanup job and startup sweep remove expired files.
Billing evidence
A billable response carries x-request-id. The same ID is used for hold settlement and the usage ledger, so one charge can be reconciled to one call. Historical usage is not repriced after a rate change.
Upstream boundary
Model output depends on upstream accounts and services outside NapKey's direct control. Pooling and failover reduce individual account failures but do not create a guarantee of uninterrupted availability.
Current limits
NapKey does not currently offer a formal SLA, tax invoices or bank withdrawals of wallet balance. Public status reports current reachability, not historical uptime guarantees.