Pay for ingest rate, not per event.
No surprise per-request bill at the end of the month. You commit to an events-per-second ceiling and we hold to it. Cross it and the SDK starts dropping events locally — you never get a runaway invoice.
Loading pricing…
One request event per top-level SDK call (e.g. GetBalance, PushTx) plus any score-change events your
scoring engine emits. Internal retries on the same provider don't multiply the count —
that's the per-attempt trace, not separate events. You can sample below 100% with the SampleRate option.
The cloud agent has a bounded ring buffer; on overflow it drops oldest events and bumps a counter. The SDK call site never blocks. Your blockchain ops keep running. The dashboard surfaces the drop count + a contextual upgrade prompt so you see the ceiling before it bites.