Feature-level margins
See whether each AI feature is accretive or margin-negative at current adoption.
Opsmeter.io
AI Cost & Inference Control
Vercel AI SDK cost tracking
Vercel AI SDK cost attribution
App-level AI SDK usage grows fast. Opsmeter maps endpoint and feature attribution so teams can control spend before rollout debt compounds.
Where teams usually miss visibility
SDK-level usage can mask which endpoint or feature is causing higher cost per request across tenants.
Release safety
Catch prompt and model changes that inflate spend before customer-wide rollout.
Operational ownership
Route spend anomalies to the feature team that can fix them immediately.
How Opsmeter tracks this dimension
Pass endpoint and feature tags
Attach endpointTag and feature keys alongside request IDs in SDK calls.
Normalize telemetry payloads
Opsmeter aligns request usage into a schema built for cross-feature analysis.
Open concentrated cost views
Group by endpoint, feature, and tenant to identify true cost ownership.
Feature attribution snapshot
| Dimension row | Metric | Context |
|---|---|---|
| feature: auto-summary | $0.048 / request | Output token growth after prompt update. |
| endpoint: /api/chat/support | $1,340 / month | Highest endpoint spend concentration. |
| tenant: enterprise-gamma | 27% of feature cost | Heavy adoption from one customer segment. |
Operationalize this attribution model
Start with one endpoint, then expand attribution coverage by tenant, user, and workflow.