Where to see Claude Code spend: the screen for each setup

Claude Code spend appears on a different screen for each of the three ways a company gets the tool, and a manager who opens the wrong one sees nothing, or sees an estimate and takes it for an invoice. This page lists each screen, what it shows and how often it changes, from Anthropic's documentation read on 4 October 2026. It also explains the one figure a developer can show you on their own machine, and the setting that makes that figure match a contract rate. It is part of the guide to Claude Code costs for teams; the three setups themselves are explained on the page on subscription seats or pay per token.

A few terms first. A token is a piece of text the model reads or writes, and all Claude Code usage is counted in tokens. A dashboard is a web page that shows figures and charts from an account. A CSV file is a plain spreadsheet file, one row per record, that opens in Excel or Google Sheets. An API is the interface that programs use to send requests to a service.

Teams and Enterprise plans: two screens in the claude.ai admin console

On a Claude for Teams or Enterprise plan, each member's usage draws from a seat allowance, an amount of usage included in the seat, and Anthropic's documentation says usage inside that allowance "isn't metered in dollars" [1]. Dollars appear only for usage credits, which is paid usage a member draws on after the allowance runs out. That fact decides what the two screens can show.

The spend report in organisation analytics. Anthropic's documentation says this report "shows estimated spend per user and per model, with CSV export, updated daily", that it covers usage-credit spend, and that it "appears once usage credits are turned on" [1]. So an organisation that has never turned on usage credits has no spend report, because there is no dollar spend to report. The Teams and Enterprise analytics documentation adds that the export lists token usage and estimated usage-credit spend per user and per model [2].

The analytics dashboard at claude.ai/analytics/claude-code shows adoption. Anthropic's analytics documentation, read on 4 October 2026, says Admins and Owners can view it, and that it includes usage metrics (lines of code accepted, suggestion accept rate, daily active users and sessions), contribution metrics (pull requests and lines of code shipped with Claude Code assistance, which need a GitHub integration), a leaderboard of the top 10 users, and a CSV export of contribution data for all users [2]. Usage and adoption data is available for all Teams and Enterprise accounts; contribution metrics need an Owner to enable them and a GitHub administrator to install the GitHub app, and Anthropic says the data "typically appears within 24 hours after enabling, with daily updates" [2]. Organisations with Zero Data Retention enabled see usage metrics only [2].

Put together: on a subscription plan, the dashboard tells you who is using the tool and how much they ship, and the spend report tells you what usage credits cost, per person and per model, as of the previous day.

The Claude Console: the usage page and the dashboard

The Claude Console is Anthropic's web console for accounts billed per token through the API. Anthropic's documentation names two screens.

The Console usage page at platform.claude.com/usage is the billing record. Anthropic's cost documentation calls it the place for "authoritative billing" when it explains that every figure Claude Code shows a developer is an estimate [1]. Spend is organised by workspace, a named compartment of the account, and a workspace called "Claude Code" is created automatically the first time Claude Code authenticates with a Console account, so that all Claude Code usage in the organisation is tracked in one place [1].

The Console dashboard at platform.claude.com/claude-code shows, by Anthropic's analytics documentation: lines of code accepted, suggestion accept rate, an Activity chart of daily active users and sessions, and a Spend chart of "daily API costs in dollars alongside user count" [2]. Below the charts, a team insights table lists every user who has authenticated to Claude Code, with their spend this month and their accepted lines this month; users who signed in with an API key appear by key identifier, and users who signed in with an account login (OAuth) appear by email address [2]. Anthropic adds a note: "Spend figures in the Console dashboard are estimates for analytics purposes. For actual costs, refer to your billing page" [2].

Access to the dashboard needs the UsageView permission, which Anthropic says is granted to the Developer, Billing, Admin, Owner and Primary Owner roles [2]. The dashboard shows usage and spend metrics; the GitHub contribution metrics belong to the Teams and Enterprise dashboard [2].

Cloud providers: the cloud billing console

On Amazon Bedrock, Google Cloud's Agent Platform and Microsoft Foundry, Anthropic's documentation says Claude Code "is billed per token to your cloud account, and spend controls live in your cloud provider's billing console" [1]. The same page says Claude Code "does not send metrics from your cloud back to Anthropic", so the analytics dashboards above and the Claude Code Analytics API do not cover this usage [1]. The only Anthropic-provided screen for a cloud-provider deployment is the one on each developer's own machine, described next, plus whatever your company builds from OpenTelemetry export, a standard way for each developer's machine to send usage measurements to a monitoring system the company runs, described on the page on per-developer usage reporting.

The figure on a developer's machine: /usage is a local estimate at list price

A developer can type /usage inside Claude Code and show you a Session block with a line such as Total cost: $0.55 and a breakdown of input, output, cache read and cache write tokens per model [1]. Four things about that figure matter to a manager, all from Anthropic's cost documentation read on 4 October 2026.

First, it is computed on the developer's machine from token counts, at list price, "unless a modelPricing table is in effect" [1]. Second, Anthropic calls it an estimate and sends the reader to the Console usage page for billing [1]. Third, it covers the current session: the totals reset when /clear starts a new session, and before version 2.1.211 they kept accumulating for the life of the process [1]. Fourth, the Session block is, in Anthropic's words, "intended for API users"; on a subscription the session cost figure "isn't relevant for billing purposes", and subscribers see plan usage bars, activity statistics and a usage breakdown on the same screen instead [1]. Those plan figures are, Anthropic says, approximate and computed from local session history on that one machine, so usage from other devices and from claude.ai is left out [1].

The same session total appears in the status line's cost field, the line at the bottom of the terminal that a developer can configure [1]. The sister guide's page on how to read /usage and /context explains every line of that screen for a developer.

Reporting at your contracted rate: the modelPricing setting

If your company pays contracted rates instead of list price, every figure Claude Code shows a developer differs from the bill by the size of the discount. Anthropic's documentation says so directly: by default Claude Code computes every cost figure at list price, "so if your organization pays contracted rates, the figures in /usage, the status line, and OpenTelemetry don't match your bill" [1]. The fix is a setting named modelPricing. It changes the figures Claude Code reports. The invoice stays as the contract sets it [1].

The setting is a managed setting: a setting an organisation deploys to every developer's machine that no user, project or local setting can override. Anthropic's settings reference, read on 4 October 2026, gives its rules [3]:

• It requires Claude Code version 2.1.242 or later.
• It applies to /usage, the status line, the Agent SDK's total_cost_usd, the --max-budget-usd limit, and the OpenTelemetry cost metric and events.
• It has two optional fields. multiplier scales every cost figure: below 1 for a discount, above 1 and up to 10 for a markup. A markup requires version 2.1.271 or later; earlier versions ignore a multiplier above 1 with a warning. overrides lists, per model, four rates in US dollars per million tokens: input, output, cacheRead and cacheWrite, all four required, with cacheWrite covering both the five-minute and the one-hour cache write.
• Claude Code uses the rates exactly as written. An administrator types them in from the contract and updates the setting when the contract changes, because Claude Code reads nothing from the contract or the Claude Console.
• Its scope is managed only. Claude Code ignores the key in user, project and local settings, in the --settings flag, and on Windows in the user-writable HKCU registry.

Managed settings reach a machine in one of four ways, by Anthropic's managed settings documentation: server-managed settings set in the claude.ai admin console or on a self-hosted Claude apps gateway (a server the company runs between Claude Code and the model provider), fetched at startup and polled hourly; a device management (MDM) profile or operating system policy, read at startup and checked every 30 minutes; a managed-settings.json file in a system directory, reloaded when the file changes; or a policy helper program [5][3]. With server-managed settings, each session reports at list price until its settings fetch has confirmed the setting [3].

To confirm the rates are in effect, a developer runs /usage in a session that has received the managed settings: the Session block's Total cost line then shows the note at your organization's configured rates [1]. Anthropic adds two limits: the figures are still estimates, and the per-million-token prices in the /model picker stay at list price [1].

Every screen in one table

Setup	Screen	What it shows	How often it changes
Teams or Enterprise	Spend report in organisation analytics	Estimated usage-credit spend per user and per model, CSV export; appears once usage credits are on [1]	Updated daily [1]
Teams or Enterprise	Analytics dashboard, claude.ai/analytics/claude-code	Daily active users, sessions, lines accepted, accept rate; contribution metrics with GitHub; leaderboard; CSV export [2]	Contribution data typically appears within 24 hours of enabling, with daily updates, by Anthropic's wording [2]
Claude Console	Usage page, platform.claude.com/usage	Billing record for the account and the Claude Code workspace [1]	The documentation states no interval
Claude Console	Dashboard, platform.claude.com/claude-code	Daily API costs in dollars with user count; per-member spend this month and lines this month; estimates [2]	The documentation states no interval
Cloud provider	The cloud provider's billing console	Per-token charges to the cloud account [1]	Set by the cloud provider
Any setup	OpenTelemetry export to your own monitoring system	Per-user token and cost metrics, which Anthropic calls approximations [4]	Default export interval 60,000 milliseconds, which is one minute [4]
Any setup, one developer	/usage inside Claude Code	Session token counts and an estimated cost at list price, or at contracted rates with modelPricing [1]	Computed on the machine when the developer runs it; resets on /clear [1]

Which screen to read

Reveneau recommends one screen per purpose. For the invoice, read the Console usage page or the cloud billing console, which are the only two the documentation calls billing records. For a monthly per-person review, read the spend report on a subscription plan or the Console dashboard's team insights table on the Console, and treat both as estimates, as Anthropic does. For a figure during the month, before either screen updates, read the OpenTelemetry export if your company has set one up. And when a developer shows you /usage, read it as that session on that machine at list price, which is what Anthropic says it is. If your company pays contracted rates, deploy modelPricing before the first review, so that the figures developers see and the figures you see are computed at the same rates. The page on spend limits for an organisation, a group and a person covers what to do when a figure is too high.

Reveneau is an AI software development consultancy. All of its code is written by AI, so token use is a running cost of every Reveneau build, and Reveneau is independent of Anthropic. Every screen name and version number on this page is from Anthropic's documentation read on 4 October 2026. If you want help setting up spend reporting for a team, contact Reveneau.