Anthropic released the first beta of its Enterprise Analytics API this week. Admins can pull token usage and dollar cost through a programmatic endpoint, broken down by user, model, context window, region, and product surface. It’s one of the most complete enterprise cost feeds we’ve seen from an inference provider.
Today, we’re shipping the CloudZero Claude Enterprise adapter built on that API. It brings Claude spend into the same allocation model customers already use for the rest of their stack.
Customers have been asking for this kind of visibility for months. We work closely with Anthropic as a partner, so when the beta landed, we were ready.
Why this matters
Until now, Claude Enterprise was a partial black box for finance and engineering leaders. Centralized billing simplified procurement and security, but it traded away the per-user and per-team visibility teams used to get from individual seats. Aggregate token counts don’t tell you which teams are driving spend.
Anthropic’s new API changes that. Enterprise admins can now query usage by user, model, context window, region, and surface. That’s enough data to allocate spend cleanly across the business.
From consumption to team accountability
Consumption dimensions tell you what was used, not who in your business used it. Allocation is the work of mapping that usage back to teams, budgets, and cost centers. That’s the layer CloudZero adds.
With the adapter live, customers can run Anthropic’s user-level data through CloudZero’s allocation engine and answer the questions finance and engineering leaders are bringing to AI spend:
- “Engineering, Sales, and Customer Success all have Claude seats. I need to know how spend is splitting across them, and whether the distribution matches what we’re paying each team to do.”
- “Are we paying premium-model prices on tasks where a smaller model would do? Which teams are driving the mix, and what would tuning it save us?”
- “Each department has a Claude budget this quarter. I need to see who’s tracking, who’s over, and who’s burning their allocation faster than expected.”
Allocation is what gets these questions answered. A $1,200 user-level charge from Anthropic shows up in CloudZero against the Customer Success team’s $40K quarterly Claude budget, with three weeks left in the quarter and the team pacing 18% over plan. That’s the difference between a bill and a managed line item.
Why an allocation layer matters across the stack
There are structural reasons a dedicated allocation layer is especially valuable for AI.
Modern AI stacks are multi-provider by default. A typical deployment touches Anthropic, OpenAI, Bedrock, Vertex, and one or more self-hosted models on commodity GPUs. A spend view that only sees one of those providers is a partial picture. Nobody else is in a position to build that cross-provider view for you.
AI cost also scales differently than cloud cost. It moves with prompt size, fanout, retries, and agentic loops, where a single workflow can dwarf the spend of a hundred routine users. Finance still has to map that volatility back to teams and budgets. Otherwise AI spend stays unpredictable.
CloudZero already does this work for cloud spend. The Claude Enterprise adapter extends it into AI: Anthropic’s API delivers high-quality inference-side cost data, and our allocation engine turns it into team accountability.
Available today
If Claude is a growing line item for your organization, the CloudZero Claude Enterprise adapter is available today. We can show you Claude usage alongside the rest of your AI and cloud spend, mapped back to the teams and budgets responsible for it.
Get in touch and we’ll walk through it on your data.
FinOps In The AI Era: A Critical Recalibration
What 475 executives told us about AI and cloud efficiency.
