Claude Code reports its own usage over OpenTelemetry, so one configuration sends that data to Revenium and lands Claude Code cost in the same dashboard as your other coding assistants and providers.
This post covers how to see who used Claude Code, on what, and how costs compare across every coding assistant you run by turning on org-level metering once and pushing the configuration to developer machines, whether or not your traffic runs through a gateway.
What you're setting up
Org-level Claude Code metering runs on telemetry that travels separately from the model-call path. Claude Code exports its own usage over OpenTelemetry straight to Revenium, so setup means one configuration delivered to developer machines with nothing installed per developer.
Step 1. Pointing Claude Code at Revenium
A set of OTEL_* environment variables point Claude Code at Revenium's OTLP endpoint, authenticated with a Revenium metering key, and Revenium's Claude Code setup guide lists each one. Usage events batch and export within seconds of activity, so records appear almost as they happen. You can capture usage in real time or backfill historical sessions.
Step 2. Delivering the configuration once
For teams on a direct Anthropic connection, an administrator enables telemetry one time from the Claude admin console using server-managed settings, and every authenticated developer is covered with no local setup.
If your traffic runs through an LLM gateway such as LiteLLM, the admin console push won't reach clients, because it needs a direct organization session with Anthropic. The same variables travel through device management instead, using Jamf, Intune, or Workspace ONE, following Anthropic's MDM examples. Metering still works either way, because the telemetry rides its own connection and the route your inference takes doesn't change capture.
Step 3. Seeing everything in one dashboard
Once telemetry is flowing, Claude Code cost appears next to Cursor, Codex, Gemini CLI, and your other providers in one view. Teams attribute spend by developer and set it against merged pull requests, with anomaly alerts firing before the invoice arrives.
Client telemetry or gateway metering
If you also meter at the gateway, choose a single metered capture path for each workload. Client telemetry and in-proxy metering both observe the same calls independently. Client telemetry is the right system of record for Claude Code, because it carries the per-user session detail a proxy can't reconstruct, right down to which skills ran.
Why this beats a renewed seat license
A seat license renewed once and forgotten won't show you a single hour of real consumption. Org-level metering shows you who used Claude Code and how that cost compares against every other coding assistant you run.
You can find the full setup steps for Revenium's AI coding data reference here: https://docs.revenium.io/track-and-control-costs/analyze-ai-tooling-spend/ai-coding-data-reference
Further reading
- Claude Code monitoring reference, the metrics, events, and attributes Claude Code emits over OpenTelemetry
- Claude Code network config, which calls still reach api.anthropic.com when inference runs through a gateway
- Revenium budgets and alerts, the metered and billed cost basis for guardrails
- Revenium cost controls, how enforcement and the rule cache behave
- Revenium provider integrations, gateway-side coverage for everything else routed through the proxy
- LiteLLM unified endpoint, the /v1/messages route Claude Code speaks through a gateway
Not yet using Revenium? Sign up for free!
