Claude platform becomes generally available on AWS

Anthropic just changed how large companies can buy and use Claude on AWS. The important part is not “Claude is on AWS” — that was already true through Amazon Bedrock. The change is that AWS customers can now use Anthropic’s own full platform experience, with AWS identity, AWS billing, and AWS controls, without setting up a separate Anthropic commercial relationship. That sounds like plumbing. But in enterprise AI, plumbing is often the whole game. ### Wasn’t Claude already on AWS? Yes — but in a narrower way. Claude models have been available in Amazon Bedrock, where AWS acts as the managed model layer. What became generally available on May 11, 2026 is “Claude Platform on AWS,” which AWS describes as direct access to Anthropic’s native Claude Platform through an AWS account. AWS also says it is the first cloud provider to offer Anthropic’s native platform experience this way. (aws.amazon.com) ### What’s actually new here? The new piece is the full platform surface. AWS says customers get the same APIs, console, and developer features they would get from Anthropic directly, but tied into AWS authentication and billing. That includes the Claude Console for prompt development and evaluation, plus platform tools like web search, web fetch, code execution, prompt caching, citations, batch processing, and the Files API in beta. (aws.amazon.com) ### Why do IAM and billing matter so much? Because enterprise rollouts usually stall on account sprawl and procurement. If a team wants to test a new AI tool, security wants identity controls, finance wants one invoice, and procurement wants fewer vendor reviews. AWS says Claude Platform on AWS removes the need for separate credentials, contracts, or billing relationships. So a company that already trusts AWS IAM and tagging policies can pilot Claude with the same control structure it already uses elsewhere. (aws.amazon.com) ### What features come with it? Quite a lot, and that is why this launch is more than a reseller arrangement. AWS lists Claude Managed Agents as beta, along with advisor strategy in beta, Skills in beta, the MCP connector in beta, and the Files API in beta. It also includes web search, web fetch, code execution, prompt caching, citations, batch processing, and the console. In other words, Anthropic is exposing the agent-building stack, not just raw model inference. (aws.amazon.com) ### What are Managed Agents, exactly? Anthropic’s framing is useful here. Managed Agents separate the “brain” from the “hands” — the model decides what to do, while Anthropic manages the session, harness, and sandbox that actually run tools and code. Basically, it is Anthropic trying to make long-running agents feel more like a cloud service than a pile of custom orchestration code. That matters for teams building coding agents, research agents, or internal workflow bots. (aws.amazon.com) ### How does this fit Anthropic’s broader push? It lines up with Anthropic’s bigger move toward agentic developer workflows. Over the past year, Anthropic has expanded Claude Code, added code execution, the Files API, MCP connectivity, longer prompt caching, and stronger memory behavior in newer Claude models. Claude Platform on AWS gives enterprises a cleaner way to adopt that whole stack inside an environment they already govern. (anthropic.com) ### So why does this matter now? Because the AI platform fight is shifting from “whose model is best?” to “which stack is easiest to deploy safely at scale?” Bedrock already gave AWS customers access to Claude models. This new launch goes after the next bottleneck — operational friction. If Anthropic can make its native platform feel like just another AWS service, it gets closer to becoming default infrastructure inside big companies, not just a model vendor on a menu. (anthropic.com) ### Bottom line? This is Anthropic and AWS trying to erase the seam between cloud infrastructure and AI application layer. The models mattered first. Now the control plane matters more. And that is exactly where enterprise buying decisions get made. (aws.amazon.com 1) (aws.amazon.com 2)