Anthropic took the Claude Platform on AWS out of preview and into general availability, giving AWS customers a way to consume the full native Claude API—every feature, every beta, every new model—on day one through their existing AWS account. Authentication runs through IAM, audit logging through CloudTrail, and the entire bill rolls up onto a single AWS invoice that retires existing AWS commitments. Claude on Amazon Bedrock is not going away; this is a different path with different trade-offs.
Two Paths, One Choice
Until now, AWS customers had one way to use Claude on AWS: through Amazon Bedrock, where AWS is the data processor and the feature set is whatever Bedrock has wired up. That model still exists, and it's still the right choice for teams with strict regional data residency requirements that need data processed exclusively inside the AWS boundary.
The Claude Platform on AWS sits on the other side of the trade-off. Anthropic operates the service, data is processed outside the AWS boundary, and in exchange you get every native Claude API feature the same day it ships—including the betas. The headline framing from Anthropic: "a first of its kind offering for Anthropic, giving you all native Claude API features from day one."
What's in the Box
The platform ships with the full native Claude stack: Claude Managed Agents (beta) for deploying agents at scale, the Advisor strategy (beta) that lets agents consult an advisor model for an intelligence boost, web search, web fetch, code execution for running Python and rendering visualizations inside an API call, the Files API, Skills, the MCP connector for hitting any remote MCP server without writing client code, prompt caching, citations, and batch processing for high-volume async workloads.
It also includes the Claude Console—Anthropic's development environment with a prompt improver, prompt generator, and evaluation tools—accessible to every Claude Platform on AWS customer. Claude Opus 4.7, Sonnet 4.6, and Haiku 4.5 are available at launch, and Anthropic is committing that new models will ship on the AWS path as they launch on the native API.
How the Plumbing Actually Works
The IAM-and-CloudTrail integration is the part that matters for security teams. Customers use their existing AWS credentials and IAM policies to authenticate against the Claude API, which means no new key vault, no separate Anthropic credential to rotate, and no parallel access-control matrix. CloudTrail picks up the audit log so SOC and compliance teams keep using the same tooling they already have. Billing flows through AWS the same way: a single invoice, full commitment retirement against existing AWS spend, EDP and PPA discounts where applicable.
Availability targets most AWS commercial regions, with both global and U.S. inference geographies supported. Anthropic flagged one operational note for existing Bedrock customers: if you have a Bedrock private offer in flight, contact your Anthropic or AWS account executive before switching, because discounts can't be applied retroactively to usage incurred before the Claude Platform private offer is accepted.
Why It Matters for Web Developers
For teams already deep in AWS, this collapses two of the biggest reasons to wait on adopting Claude features in production. The "we'll get it when Bedrock catches up" gap is gone—the moment Anthropic ships a new model or capability on platform.claude.com, it's also live on AWS through the same invoice your CFO already approves. And the procurement story stays clean: no new vendor contract, no separate spend center, no renegotiation of the AWS commitment.
The endorsements Anthropic shipped with the announcement underline the use case. OpenRouter is using it because IAM gives them the same access controls they already use for everything else AWS. Carlyle framed it as "the canonical Anthropic API with AWS as the access layer." For platform teams building agents on Claude Managed Agents, the AWS path is now the cleanest enterprise on-ramp Anthropic has shipped—available today, with parity guaranteed going forward.