How Kadryn works
Understand the main Kadryn data and control flow.
The operating loop
Kadryn follows a simple operating loop: collect usage, normalize cost, attach context, detect issues, enforce policies and recommend improvements.
That loop lets teams answer operational questions quickly. Which feature caused the spike? Which project is missing metadata? Which request was blocked by a policy? Which model substitution would reduce cost without hurting quality?
Data enters Kadryn
Usage enters Kadryn from gateway traffic, direct ingestion events, provider integrations, Slack workflows and webhooks. Each source should include enough metadata to identify the owning project, team, feature and environment.
Kadryn normalizes usage
Providers expose usage and pricing differently. Kadryn converts provider-specific usage into normalized records so costs can be compared across models, vendors and product surfaces.
Teams act on the data
Once data is normalized, Kadryn powers dashboards, alerts, guardrails, approvals, incidents, optimization opportunities and finance reports.
What can go wrong
Most setup issues are caused by missing metadata, client-side secrets, inconsistent environment names or usage events that cannot be matched to a project. Fix those early before enabling enforcement.