Anthropic's new Usage and Cost API reports Claude spend by user and key

Anthropic announced on 29 October 2026 that its Usage and Cost Analytics API is available in beta. According to the company's release notes, two endpoints, /v1/organizations/analytics/usage and /v1/organizations/analytics/cost, report token usage and cost for Claude Console and Claude Enterprise organizations. You can filter and group results by model, workspace, API key, and the user or service account behind each request.

What it lets a finance team ask

Anthropic's documentation describes four uses: attributing spend, finding top consumers, building time series, and using one set of endpoints for both Console and Enterprise organizations. The example it gives is ranking which API keys spend the most in a month. The cost endpoint reports an amount after discounts and a list_amount at list prices, in USD, so a team can see what a negotiated discount is worth. It can also split cost by cost type and token type, including cached input tokens.

• A query takes up to five grouping dimensions, and up to 100 values per filter.
• Buckets can be one minute, one hour or one day; a request can span at most 31 days, and one-minute buckets at most 24 hours.
• Start dates must be on or after 1 January 2026 and within the last 365 days.
• Cost amounts are decimal strings in cents with six decimal places: Anthropic's example "41280.000000" is $412.80.

What the documentation says to watch

Timing is the main catch. Anthropic says data typically refreshes within 4 hours and can take up to 24 hours, and values can be revised until about 7 days after the end of the calendar month. For final totals, it advises querying only months that ended at least 7 days ago. The legacy API's data typically appears within 5 minutes, per Anthropic's migration table, so a dashboard built for same-hour alerts will not get that from the new one.

Requests served under a Priority Tier commitment show up in usage with service_tier "priority" but add zero token cost, because the commitment covers them. The commitment fee is billed separately and is not in this data, so a cost report built only on this API will understate that spend. Access also needs the right credential: an Admin API key for Console organizations, or a key with the read:analytics scope created in claude.ai for Enterprise organizations.

Migration

The release notes say the older Usage and Cost API, and the usage_report, cost_report, user_usage_report and user_cost_report endpoints of the Claude Enterprise Analytics API, are now legacy but remain supported. No removal date is stated in the notes I read. Anthropic lists changes that will break a naive port: the beta header is required, rows come back ranked rather than as a list of time buckets, and combinations with no usage have no row, so a daily chart must fill missing days with zeros. Cost tiers and amount formatting also change, and workspace_id is now the default workspace's real ID instead of null.

What to do with it

If a CFO asks what AI costs per team, start by tagging each team or project with its own workspace or API key, then group cost by workspace and model for a closed month and compare it with the invoice. Treat the result as a monthly allocation report, not a live meter. Keep the legacy API for any alert that needs sub-hour data until you have tested the gap.