Who it’s for · insurance teams
Design the trigger.
See every assumption.
For the program designers and underwriting teams whose job already is these verbs: schedule in, trigger as code, forty seasons replayed, a price with its assumptions on the face of it.
Get a free API key → Start with the quickstart →
Backdrop: Super Typhoon Vongfong, Philippine Sea, 9 October 2014 — Aqua/MODIS, NASA, public domain.
The job, end to end
From the file you were sent to a price you can defend
-
Ingest the schedule as it arrived
Merged cells, header drift, mixed encodings, ambiguous units — ingest is measured against files in exactly that state, and writes every guess it made into
source_fidelityinstead of hiding it. -
Write the trigger as code
An index over a pinned feed, thresholds, a payout function — fields a machine validates, not prose a lawyer argues about later. The trigger object is the reference.
-
Replay it across forty seasons
The backtest runs against the pinned record year by year and shows which occurrences paid and at what fraction of limit — descriptive history with thin data years disclosed, never a forecast.
-
Read the technical price with its working
Model versions, data vintages, event set, loadings and currency basis arrive as fields beside the number. What you do with the number is your business — Craton outputs the analytics and stops there.
Receipts, not references
What you can check before you trust it
The honesty exhibit
What ingest survives, flags and refuses, measured on the cursed corpus and regenerated from the manifests — never typed by hand.
One run, published whole
A real price run with its run id, assumptions and vintages, and the block that reproduces its digest byte for byte.
The measured walks
How long onboarding actually takes, timed by the production checks on every sweep and rendered from the log.