Claude Enterprise gets model-level analytics, role-based entitlements, and spend alerts
Anthropic adds per-model usage analytics, admin-configurable model entitlements, and spend-threshold alerts to Claude Enterprise.
Update, 3 July 2026: Claude Enterprise gets model-level analytics, role-based entitlements, and spend alerts — already covered
This update confirms no new information. The story published here on 2026-07-02 is the same announcement covered in Anthropic’s official blog post, and Anthropic has issued no corrections or additions since.
The original article stands as accurate. One detail worth flagging that was not prominently surfaced in the original framing: individual member-level usage analytics is currently off by default, but Anthropic has confirmed it will switch to on by default starting July 11, 2026. Organizations that want to keep member-level visibility disabled must actively change the setting on or after that date. The Claude Help Center article on usage analytics has the steps for doing so.
Everything else in the original post remains current: per-model cost breakdowns in the analytics dashboard, the Analytics API for pulling data into third-party tools, admin-configurable model defaults and role-based entitlements, spend-threshold alerts at 75% and 90% for admins, and the Admin API for managing controls at scale. No corrections to pricing, feature scope, or availability have been issued.
Managing Claude in a mid-to-large organisation has been a bit of a black box. You could set a spend cap, but knowing which teams were burning through tokens, which models they were using, and whether anyone was about to hit a wall mid-task required either exporting CSVs manually or piecing it together from the Analytics API yourself. Anthropic has now addressed that directly with three additions to Claude Enterprise: model-level usage analytics, per-role model entitlements, and configurable spend-alert thresholds.
These sit on top of existing controls rather than replacing them, and they’re clearly aimed at organisations running Claude at scale, particularly where agentic workloads are involved.
What’s actually changed
Usage analytics that go deeper than “how much did we spend”
The admin dashboard now shows cost and usage broken down by group and by individual user. Crucially, it surfaces what those users are doing: artifacts created, files edited, skills and connectors invoked, all displayed alongside their cost contribution. That shift from “total spend” to “spend by activity type” matters when you’re trying to justify AI costs to finance or understand which workflows are actually getting traction.
For teams that want to pull this into their existing tooling, the Analytics API now supports filtering by date range, team, product, or model. Integrations with Datadog Cloud Cost Management and CloudZero are available, so you can put Claude usage next to the rest of your cloud spend rather than managing it in a separate tab. Skills report their own usage and cost independently, and new endpoints track plugin adoption and artifact creation.
One limitation worth knowing: the Analytics API only tracks Claude Code sessions running against the Claude API directly. Sessions routed through Bedrock, Vertex, or Microsoft Foundry don’t appear in it. For enterprises using those channels at scale, that’s a meaningful gap in reporting coverage.
Model entitlements admins can actually configure
Previously, new conversations defaulted to whichever model Anthropic had set as the standard. Now admins can set which model new conversations start with across Claude chat, Cowork, and Claude Code, and they can control which models are available to specific roles or across the whole organisation.
The practical value here is straightforward. Most routine work doesn’t need Opus. If your customer support team is doing simple summarisation and your developers are doing complex multi-step reasoning, those shouldn’t be defaulting to the same model. Being able to route by role means you’re not subsidising premium compute for tasks that don’t require it.
Spend alerts before people get blocked
The spend-threshold alerts are arguably the most immediately useful addition for day-to-day admin life. Admins now get notified at 75% and 90% of an org-level spend limit, which gives enough runway to raise the cap before anyone actually gets cut off. Users get their own in-app notifications at 75% and 95% of their individual limit and can request a limit increase directly from inside Claude without needing to file a ticket.
This addresses a specific and frustrating failure mode: someone’s mid-task on something important, they hit their cap, and everything stops. The alert structure is designed to prevent that from being a surprise.
For organisations managing limits across many groups, the Admin API extends this to automation. You can script increase-request reviews, flag users approaching their limits, and identify rapidly changing usage patterns without doing it manually per team.
What this means for IT and FinOps teams
The honest summary is that Anthropic is catching up to where enterprise software governance expectations already were. The tools being added here, per-model cost visibility, role-based access controls, proactive spend alerts, are things IT and FinOps teams would expect from any mature SaaS platform they’re managing at scale.
What makes it timely is the shift toward agentic workloads. When Claude is autonomously browsing, writing code, editing files, and calling APIs over extended sessions, a single workflow can consume tokens at a rate that has no comparison to a user typing questions into a chat window. Cost patterns become harder to predict, and the blast radius of an uncapped spend limit gets larger. The controls Anthropic is adding here are sized for that reality, not just for chat.
If your organisation already has Claude Enterprise and you haven’t looked at the Analytics API or the admin dashboard recently, it’s worth revisiting. The programmatic access in particular, the ability to pull per-user, per-model, per-product cost data into Datadog or your own internal tooling, closes a reporting gap that was pushing some teams toward spreadsheet-based workarounds.
A few things to check if you’re running Claude Enterprise
- Review your model defaults. If every role in your org is defaulting to the most capable model, you may be paying for capability your workflows don’t need.
- Set up spend alerts if you haven’t. The 75%/90% thresholds for admins and 75%/95% for users are configurable and opt-in.
- Check your Analytics API coverage. If you’re running Claude Code through Bedrock or Foundry, those sessions aren’t captured in the Analytics API yet. Factor that into any usage reporting you’re building.
- Explore the Admin API for scale. If you’re managing spend limits across dozens of teams, scripting that through the Admin API will save significant manual overhead.
The full details of the update are covered in Anthropic’s admin visibility and controls post. The Analytics API documentation and endpoint reference are available through the Anthropic platform docs.