To see who spent the AI budget, use the reporting that matches how you buy: seat reports for subscriptions, workspaces for the API, your cloud console for cloud hosted models, and OpenTelemetry or a gateway when you want per user numbers in your own tools. Pick one source of truth per setup.
Match the method to the purchase
Anthropic's Claude Code cost documentation maps three setups to where you see spend, where you cap it and how you get per user data. As of 19 October 2026:
| Setup | See spend | Cap spend |
|---|---|---|
| Teams or Enterprise seats | Spend report in organization analytics | Spend limits in admin settings |
| Console API | Console usage page | Workspace spend limits |
| Amazon Bedrock, Google Cloud or Microsoft Foundry | Your cloud billing console | Your cloud's budget controls |
On seat plans, the page says usage inside the seat allowance is not metered in dollars, and the spend report covers usage credit spend once credits are turned on.
Workspaces as cost centers
For API use, create a workspace per team, per product or per environment, and set a spend limit on each. The page notes that when Claude Code first authenticates with a Console account, a workspace called Claude Code is created automatically for centralized tracking, and that you can set a rate limit on it to protect other production workloads.
Per user numbers in your own stack
OpenTelemetry export works on every setup and is described as the only option that streams per user token and cost metrics into your own observability stack in near real time. A gateway that routes all traffic through a proxy can track spend per key, and the page names an open source example while noting it is unaffiliated with Anthropic and not audited for security. If you add a gateway, treat it as a sensitive system: it holds your keys and sees your prompts.
Make the numbers match the invoice
By default, tools compute cost at list price. The page says Claude Code's dollar figure is calculated locally at list price, and offers a managed setting so administrators can enter contracted rates. It also says the figure is an estimate and the Console usage page is authoritative for billing. If your showback report disagrees with the invoice, this is the first place to look.
A showback checklist
- Tag every key or workspace with an owner and a product.
- Export usage monthly into one sheet, with the price date next to it.
- Report cost per team and cost per outcome, such as per merged pull request or per resolved ticket.
- Review the top five users or jobs with their managers.
- Set limits at team level, and write down who can raise them.
Pair this with a pilot method from budgeting AI coding tools for a team, and read enterprise LLM budget management patterns for wider governance.
Quick check
Is showback the same as chargeback?
No. Showback reports costs to teams. Chargeback moves the money. Start with showback.
Can I trust per user dollar figures from a coding tool?
Treat them as estimates. Reconcile them against the provider invoice or console each month.
Sources
Last reviewed: 19 October 2026.
