SDKs
PayInference is an HTTP API first — every SDK is a thin, optional convenience
over the same two calls (POST /v1/decision, POST /v1/outcomes). Nothing in
the product requires one, and the API reference is the
contract of record.
Available#
| SDK | Package | Use it for |
|---|---|---|
| Node.js | @payinference/sdk |
Server-side decisions, outcome reporting, attempt telemetry |
| Browser helpers | @payinference/sdk-js |
Checkout-page helpers only — no authentication, no API access |
More languages are planned. Until one exists for your stack, call the API
directly: it is plain JSON over HTTPS with an x-api-key header, and the
API reference documents every endpoint and field.
Which one do I need?#
Almost certainly the Node.js SDK, on your backend. The API key is a secret, so decisions are always requested server-side.
@payinference/sdk-js is deliberately not a client for the API — it holds no
credentials and cannot reach PayInference. It exists so the checkout page can
render what your backend already decided.
What an SDK gives you over raw HTTP#
- Request validation before anything leaves your process, so card data and unknown fields are rejected locally rather than by the server.
- A hard latency budget on
decide(), separate from the longer budget used for reporting calls. - An optional local fallback decision when PayInference is unreachable, so a checkout never blocks on an outage.
- Idempotent retries on reporting calls.
All of it is reproducible over plain HTTP — see Idempotency and retries and Production integration.