Anthropic ships full Claude Platform on AWS — what's different from Bedrock?

Anthropic has announced Claude Platform on AWS as Generally Available. One-line summary: AWS customers can now access Claude API's full feature set directly from Anthropic, with AWS-native auth + billing — not via Amazon Bedrock. Bedrock stays in parallel, but with this release its role shifts from "feature-rich primary path" to "data-residency fallback."

What was announced

Claude Platform on AWS is a distinct product that serves the full Anthropic API over AWS infrastructure. Auth runs via IAM, audit via CloudTrail, billing via AWS commit — so the enterprise question "can I consume Claude through my existing AWS contract?" now has a clear yes. Nearly every Claude API feature is here on day zero: Managed Agents (beta), Advisor strategy, web search + fetch, code execution, Files API, Skills, MCP connector, prompt caching, citations, batch processing, and Claude Console access.

On the model side, Opus 4.7, Sonnet 4.6 and Haiku 4.5 are available; new models will land on this platform in lockstep with other distribution channels. Early adopters listed include ReliaQuest, OpenRouter and Emergent — ReliaQuest's heavy Claude usage on the security side was already public knowledge.

What changed

Until now, AWS customers had two paths to Claude: direct Anthropic API (doesn't consume AWS commit) or Amazon Bedrock (consumes AWS commit, no feature parity). Bedrock always got Anthropic's new features late — Skills, Files API, MCP connector arrived very late or are still missing. For large enterprises this was a real blocker: "I want to consume Claude via my AWS commit, but I'll lose the new features."

The new Claude Platform on AWS removes that dilemma:

Feature parity from day zero. Anything Anthropic ships lands here simultaneously. The "6 months or never" uncertainty on Bedrock is gone.

AWS-native integration preserved. IAM policies, CloudTrail audit logs, AWS billing consolidation — all intact. So the operational infrastructure you'd lose by switching to direct Anthropic API is no longer lost.

Most commercial regions are live. Global and U.S. inference geographies are supported.

Bedrock private-offer discounts can carry over — but not automatically; you'll need to contact your account executive.

Anthropic's official position: "Bedrock remains appropriate for customers with data residency requirements." Translated: Bedrock isn't dead, but it lost the feature battle. For strict regulatory cases (finance/healthcare data that can't leave a region), Bedrock's AWS-native residency guarantee still matters. But if you're starting a new integration and don't have a hard residency constraint, the default choice is now Claude Platform.

First read

I didn't test this directly in my own pipeline — as a solo developer I already consume Anthropic API + Claude Code directly, no AWS commit. So I can't measure "Bedrock vs. new platform" first-hand. But the signal from a few teams I know who consume Claude on AWS bills is clear: waiting for new features on Bedrock could take months. Skills and MCP connector still aren't fully accessible through Bedrock.

The bigger read: this is less a technical release and more a distribution strategy move. Instead of selling indirectly through Bedrock, Anthropic uses AWS infrastructure to pull customers directly onto the Anthropic platform. Three months ago they made a similar move on Google Cloud Vertex; now the AWS side is closed. The multi-cloud distribution loop is closing — Anthropic has managed to say "on every major cloud, not via the hyperscaler but directly through me."

Practical impact

For AWS-committed startups/enterprises: net upgrade. To use Skills + MCP connector + Files API you no longer have to leave Bedrock and switch to direct Anthropic. You keep your existing IAM/CloudTrail setup.

For teams consuming Claude on Bedrock: no urgent migration, but if your feature roadmap depends on a pending capability (Skills, MCP, Managed Agents) you should plan migration to the new platform now. If you have a private-offer discount, talk to your account exec before transitioning.

For solo developers + indie makers: no direct impact. The fastest path is still hitting api.anthropic.com with a key. The AWS marketplace purchase + billing benefit doesn't matter much at small scale.

For the Turkish ecosystem: there's still no AWS Türkiye region, so for TR enterprises this announcement raises the question "can I consume Claude in EU regions via my AWS commit?" EU-West is live, so EU region choice may be sufficient for residency — but KVKK compliance on the contract side needs separate review.

Limits / concerns

Three things:

Managed Agents is still beta. Anthropic's "agentic" stack has no production SLA. If you're building a critical production agent, nail down SLA + escalation path before signing.

Bedrock's future is unclear. Anthropic officially says "Bedrock continues," but the lack of day-zero feature parity makes positioning Bedrock as a long-term strategic path risky. Have a migration plan.

AWS lock-in deepens. AWS-native auth + billing integration is attractive, but it makes "exit AWS" harder. Teams with a multi-cloud strategy should price this in.

Bottom line

Three sentences: Anthropic escapes Bedrock's limitations, pulls AWS customers natively onto its own platform, and pulls this move not by competing with AWS but by embedding inside AWS infrastructure. For solo developers there's no direct impact; for AWS-committed enterprises/startups the impact is at least "re-evaluate your feature roadmap."

Singrey's note

The part that interests me most isn't the list of new features — I already get those through the Anthropic API. What I find interesting is Anthropic's strategy of using the hyperscaler's infrastructure as a channel, instead of being the hyperscaler's channel. I saw the same pattern when they integrated Sonnet 4 directly on Google Cloud Vertex. Long-term, this is AI labs trying to reclaim distribution power from hyperscalers. Compared to OpenAI's tight bond with Azure, Anthropic's multi-front distribution strategy means a healthier ecosystem for solo developers — we're not locked into a single point of dependency.