Anthropic announced a new usage-and-cost reporting beta on October 29, 2026, for organizations using Claude Console or Claude Enterprise. It gives finance and operations teams more ways to connect AI spending to people and automated services. The release notes also confirm that the older reporting tools remain supported.
For Console users, the choice is whether more complete spending breakdowns justify waiting longer for recent activity to appear. Anthropic’s beta documentation puts the usual reporting delay at up to four hours, potentially stretching to 24. That makes the new service better suited to reviewing past spending than watching a fast-moving bill.
What the fuller breakdown changes
The Usage and Cost Analytics API—a way to feed Claude data into a company’s own reports—uses a common interface for Console and Enterprise organizations. Its cost endpoint can combine dimensions such as model and attributed identity, returning every matching combination across successive pages of results.
That addresses a specific Enterprise limitation. The older grouped usage and cost reports return only their top 100 groups per time bucket; the new interface returns all groups with data. This is an improvement to the breakdown, not evidence that the old organization-wide total omitted the remaining spending. Anthropic explains the distinction in its Enterprise migration guide.
Attribution still depends on the identity recorded with a request. Some older shared keys identify the key, not an individual employee, and some activity remains unattributed. The identity rules therefore have a practical consequence: keep unassigned spending visible instead of guessing who incurred it. A per-person cost chart is not a measure of that person’s productivity.
Choose according to how quickly you need the answer
For the legacy Console API, Anthropic says usage and cost usually appear within five minutes of a completed request, though delays can be longer. Existing Enterprise cost-and-usage reporting already has the four-to-24-hour window. The beta is not introducing the same slowdown for every customer.
The beta can divide results into minute, hour or day intervals. That describes the detail in the report, not how recently it was refreshed. A minute-by-minute chart can still be missing hours of recent activity. The freshness timestamp is the relevant signal for deciding which part of a report is incomplete.
Based on those documented differences, the adoption choices are:
Your reporting need | Recommended approach |
|---|---|
Allocate past spending across many identities or categories | Evaluate the beta, particularly if Enterprise breakdowns exceed 100 groups. Retrieve every page before allocating costs. |
Watch direct Console spending where minutes matter | Retain the supported legacy feed for fresher visibility; assess the beta separately for historical reporting. |
An existing Enterprise report already answers the question | There is no immediate need to migrate. Switch for better breakdowns or shared integration code, not a presumed speed improvement. |
An occasional total already available in Console is sufficient | Do not commission an integration solely because the beta exists. First identify the decision your current view cannot support. |
Check your organization’s access and coverage
This is organization-level reporting, not a new feature for every personal Claude subscriber. Console organizations use an Admin API credential; the Admin API is unavailable to individual accounts. Enterprise access uses an Analytics API key with read:analytics permission, created by the primary owner after enabling access.
Anthropic’s Enterprise access guide says cost-and-usage endpoints apply to usage-based Enterprise plans; on seat-based plans, they reflect usage credits only. Do not interpret that as a complete account of subscription fees.
A shared interface also does not automatically combine all company accounts. Available reporting dimensions differ by organization type: Console workspace and API-key breakdowns are not universal Enterprise fields. Before commissioning a combined report, identify which organizations and spending categories it must cover.
Reconcile reported cost with the bill
The cost reference defines the reported amount as charges after discounts but before credits. It also returns amounts in fractional cents, even though the currency is USD. An integration must convert the units correctly and account separately for credits when estimating cash payable.
There is a different gap for Priority Tier commitments, where a company pays separately for committed capacity. Covered requests contribute zero token cost in these reports, while the commitment fee is excluded. That zero does not mean the capacity was free. Anthropic documents this exclusion in the beta’s cost guide.
The implication for finance is that a usage-cost total can differ from the bill in either direction: credits may reduce the cash payment, while omitted commitment fees add costs outside the report. Reconciliation needs the organization’s actual billing terms, not one universal adjustment.
Recent figures are also provisional. Anthropic says values can change until about seven days after the calendar month ends and recommends using months that ended at least seven days ago for final totals. Its freshness guidance supports a simple reporting policy: show when data was refreshed, label recent allocations provisional and rerun them after that window.
What to do before switching
Start with one reporting question your current setup cannot answer. For a proposed rollout, compare the old and new reports for the same organization, complete date range and billing basis. The migration documentation identifies changes to result ordering, pagination and empty results; an empty response should not be accepted as zero spending without checking the query.
If the real task is preventing overspend, configure the applicable controls separately. Console workspace settings provide monthly caps and threshold alerts; reading an analytics report does not set them. For a related cross-provider decision, RohitAI’s guide to Salesforce Agent Fabric’s spending controls examines which traffic those controls cover.
The beta is worth evaluating when fuller attribution changes a budget or allocation decision. Keep a working report when it already answers the question, and preserve faster Console visibility where it matters. The October 29 announcement gives no retirement deadline for the supported legacy reports.