Anthropic made Claude Platform generally available on AWS on May 11, 2026, giving companies a way to access its full native API through their existing AWS account - without switching to a separate Anthropic billing relationship.
The integration lets teams authenticate via AWS IAM, route audit logs through CloudTrail, and consolidate Claude usage into their existing AWS invoice. It also counts against existing AWS commitment spend, which matters for enterprises already locked into large AWS contracts.
What's Actually Included
This isn't a stripped-down version. The AWS-billed tier gives access to Claude Managed Agents (currently in beta - Anthropic's system for deploying multi-step AI agents), web search, Python code execution, the Files API for document uploads, MCP connectors for hooking into remote services, prompt caching (which reduces costs and response times on repeated context), citations, and batch processing for async workloads. All three current models are available: Claude Opus 4.7, Sonnet 4.6, and Haiku 4.5.
How This Differs from Claude on Amazon Bedrock
Anthropichas been clear about one key trade-off: data processed through Claude Platform on AWS runs outside the AWS network boundary - Anthropic operates the infrastructure, not Amazon. That's the opposite of Claude on Amazon Bedrock, where Amazon acts as the data processor and everything stays within the AWS boundary.
For teams in regulated industries with strict data residency requirements, Bedrock remains the safer path. For teams that need day-one access to Anthropic's latest features and want to skip the wait for Bedrock integration, the Platform route now makes sense from a procurement perspective.
Customers quoted in Anthropic's announcement cited two consistent reasons for adopting it: faster access to new capabilities (Emergent noted "day-one" feature availability) and the operational simplicity of IAM-based access control for teams already running Claude Code alongside other AWS workloads.
The service is live in most AWS commercial regions with support for both global and U.S.-only inference routing.