ConceptsstableUpdated 2026-05-14

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.