Exposing Hidden LLM Spend: Team-Based Cost Attribution Architecture

Shared LLM API keys obscure which teams are responsible for costs, making optimization impossible—a risk one team learned when a retry loop inflated spending tenfold undetected. Using OpenTelemetry and OpenInference standards to tag spans at the gateway level enables team-based cost attribution in real time, turning cost chaos into actionable insights while keeping your instrumentation portable across backends.