Claude Platform on AWS is the Anthropic-operated deployment that runs inside AWS and typically tracks the first-party Claude API feature-for-feature, same day. But "typically" carries a documented exception list — most Admin API endpoints, spend limits, and, relevant here, the Usage and Cost API. Anthropic's documentation is explicit: "Claude Platform on AWS: The programmatic Usage and Cost API endpoints are not currently available. View usage and cost data on the Usage and Cost pages in the Claude Console instead."
If your organization planned to feed a FinOps pipeline, a Datadog dashboard, or a monthly chargeback job from GET /v1/organizations/usage_report/messages and GET /v1/organizations/cost_report, and you deploy on Claude Platform on AWS, those endpoints simply aren't there to call. This article covers what you get instead and how teams typically bridge the gap.
What the Console pages actually provide
The Console dashboards are more capable than "a usage page" suggests. Per Anthropic's support documentation:
| Page | What you can see |
|---|---|
| Usage | Token breakdowns by model, date/time, and API key at hour or minute granularity; input/output token charts; filters for workspace, model, month, and API key; CSV export |
| Usage (rate-limit telemetry) | Counts of rate-limited (blocked) requests, hourly maximum uncached input tokens per minute alongside cache rates, and output tokens per minute versus your configured limits |
| Cost | Spend by model or combined, monthly breakdowns with daily visualization, totals including tool-use costs, CSV export |
The rate-limit telemetry deserves a highlight: seeing blocked request counts and your uncached input-token peaks next to cache rates is exactly the evidence you need when deciding whether to raise limits or improve prompt caching. And CSV export means the data isn't trapped — it's just not on an API.
Who can see it
Access to the Usage and Cost pages is limited to three Console roles: Developer, Billing, and Admin. Members outside those roles have no visibility, and the support documentation also notes it isn't currently possible to break usage or cost down by individual users. If your reporting audience is wider than those roles — say, engineering managers who aren't Console developers — someone with access will need to export and redistribute.
Practical workarounds
1. Export the CSVs on a schedule. The Usage and Cost pages both export CSV. A person (or a documented monthly ritual) downloading those files into your finance workflow is the lowest-effort bridge. It is manual, but the data matches what the dashboards show.
2. Capture usage client-side. Every Messages response includes a usage object with input, output, and cache token counts — and Anthropic documents that this object is consistent across platforms, including Claude Platform on AWS. Logging it per request in your own telemetry gives you programmatic usage data the platform doesn't expose, attributed however you like (user, feature, tenant). See the usage-object billing fields guide for what's in it.
3. Reconcile against the AWS bill. Billing for Claude Platform on AWS flows through AWS Marketplace, so your cloud invoice is the financial source of truth — the Console Cost page and your client-side counts should explain it.
Where to go next
If programmatic usage data is a hard requirement, compare this against the first-party Claude API, where the full Usage and Cost API is available — including its documented attribution gaps and grouping dimensions. For how billing itself flows, see billing through AWS.