Claude Platform Lands on AWS With IAM and Native Billing

Anthropic shipped Claude Platform on AWS this week, putting the full Claude API behind AWS-native billing and IAM authentication for the first time. The launch matters because it shortcuts months of procurement work for enterprises already standardized on AWS: Claude Platform on AWS bills through your AWS account, authenticates via IAM, and exposes the complete Anthropic API surface — Messages, Files, Message Batches, Claude Managed Agents, Agent Skills, code execution, and tool use — without requiring a separate Anthropic vendor relationship. This is distinct from Claude in Amazon Bedrock, which is a Bedrock-hosted model with Bedrock’s API contract; Claude Platform on AWS is Anthropic’s own platform infrastructure, running in AWS, with AWS-friendly procurement plumbing.

What’s actually new

Three things changed. First, the procurement surface: enterprises can now pay for Claude usage through AWS Marketplace and AWS billing rather than through a separate Anthropic contract. For organizations with existing AWS commitments and a finance team that strongly prefers a single cloud bill, this is a real friction remover. Second, the authentication surface: API access uses AWS IAM identities and policies, which means existing enterprise IAM controls (least-privilege roles, MFA enforcement, audit logging via CloudTrail) apply natively to Claude usage. Third, the platform surface: Claude Platform on AWS exposes Anthropic’s full API — not the Bedrock-mediated subset — so customers get Anthropic’s newest features as they ship, not after they get ported to the Bedrock contract.

The architecture matters for compliance and data-residency reasons. Customer traffic terminates in AWS regions, with AWS networking and isolation, while Anthropic operates the platform infrastructure inside those AWS environments. The deployment positions Anthropic as a co-resident enterprise vendor inside AWS rather than as a third-party API consumed across the public internet. Compliance teams that previously hesitated to use Claude over latency-sensitive or data-classification concerns get a meaningful new option.

Why Claude Platform on AWS matters

• Procurement friction drops dramatically. Enterprises with AWS Enterprise Discount Programs (EDP) and consolidated cloud spending can route Claude usage through existing AWS commitments. No separate vendor contract; no new payment terms negotiation; no new finance vendor onboarding.
• Native IAM eliminates a security-review category. AWS IAM policies, roles, and audit trails are already understood by enterprise security teams. Authentication and authorization for Claude usage now slot into the same controls as everything else in AWS.
• Feature parity with the direct Anthropic API. Claude Platform on AWS isn’t a Bedrock-flavored subset — it’s Anthropic’s full platform. Agent Skills, Managed Agents, Files API, Batches, code execution: all present from day one.
• The Claude-on-AWS ecosystem just got more confusing. Customers now have three ways to use Claude on AWS: through Bedrock (managed Bedrock service, Bedrock API), through Claude Platform on AWS (Anthropic platform, Anthropic API, AWS billing), or via direct Anthropic API calls (Anthropic platform, Anthropic API, Anthropic billing). The right choice depends on what you optimize for.
• Strategic positioning relative to Microsoft-OpenAI. Microsoft’s investment in OpenAI gives Azure customers a tight OpenAI integration; AWS’s investment in Anthropic now extends meaningfully past Bedrock with this platform launch. The cloud-AI alignment continues hardening.
• For Anthropic, AWS distribution is a serious channel. AWS Marketplace gives Anthropic access to enterprise procurement pipelines it couldn’t reach as easily before. The revenue mix between Anthropic-direct, Bedrock-mediated, and now Claude Platform on AWS will shift over the next 12 months as enterprises evaluate.

How to use Claude Platform on AWS today

The launch is live for AWS customers in supported regions. Getting started is meant to be straightforward for organizations already on AWS.

1. Confirm you’re in a supported AWS region. Anthropic launched in primary US and EU regions first; check the AWS Marketplace listing for the current region list before planning deployments.
2. Subscribe to Claude Platform on AWS via AWS Marketplace. The subscription provisions the integration into your AWS account; usage charges accrue through AWS billing from that point.
3. Create IAM roles for the principals that should access Claude. Apply least-privilege policies — Claude Platform on AWS recognizes standard IAM actions, so you can grant anthropic:CreateMessage to a specific role and limit blast radius the same way you would for any AWS service.
4. Make your first API call. The API contract is the same Anthropic Messages API you’d use against api.anthropic.com; the endpoint and authentication mechanism are what differ.
5. Wire CloudTrail logging. Every API call is recorded by CloudTrail with the IAM principal, request metadata, and response status. Use this for audit, cost attribution, and anomaly detection.

# Sample Python code calling Claude Platform on AWS

# Install Anthropic SDK with AWS auth support
pip install "anthropic[bedrock]"
# Note: actual SDK module names may vary; check Anthropic docs

# Configure AWS credentials (standard AWS chain)
# - Environment vars: AWS_ACCESS_KEY_ID, AWS_SECRET_ACCESS_KEY
# - Instance profile (if running on EC2)
# - AWS SSO / IAM Identity Center

# Example call (pseudocode pattern based on launch docs)
from anthropic import AnthropicAWS

client = AnthropicAWS(
    aws_region="us-east-1",
    # AWS credentials picked up from the default chain
)

response = client.messages.create(
    model="claude-opus-4-7",
    max_tokens=1024,
    messages=[
        {"role": "user", "content": "Summarize the value of Claude Platform on AWS in two sentences."}
    ],
)

print(response.content[0].text)

# Verify in CloudTrail that the call was logged
# AWS Console → CloudTrail → Event history
# Look for events with eventSource = "anthropic.amazonaws.com"

For organizations with strict IAM policies, define a dedicated role per workload rather than reusing a shared role. The IAM policies attached to that role become the audit trail for which workloads are using Claude and at what scale. Cost attribution follows IAM identity, so different teams running on the same AWS account can attribute their Claude spend cleanly.

# Sample IAM policy for least-privilege Claude usage
{
  "Version": "2012-10-17",
  "Statement": [
    {
      "Effect": "Allow",
      "Action": [
        "anthropic:CreateMessage",
        "anthropic:ListMessages",
        "anthropic:CreateBatch",
        "anthropic:GetBatch"
      ],
      "Resource": "*",
      "Condition": {
        "StringEquals": {
          "aws:RequestedRegion": "us-east-1"
        }
      }
    }
  ]
}

# Note: actual action names will be in Anthropic's IAM docs
# Use AWS Marketplace listing for the canonical action list

How Claude on AWS deployment options compare

The three Claude-on-AWS deployment paths each optimize for different priorities. The table below maps the differences.

Option	API Contract	Billing	Auth	Best For
Direct Anthropic API	Anthropic Messages API (full)	Anthropic invoice	Anthropic API key	Anthropic-native teams, fastest feature access
Claude in Amazon Bedrock	Bedrock API (Anthropic subset)	AWS billing	AWS IAM	Teams already deep on Bedrock, multi-model Bedrock workflows
Claude Platform on AWS	Anthropic Messages API (full)	AWS Marketplace / AWS billing	AWS IAM	AWS enterprises wanting full Anthropic features + AWS procurement

The simplest decision framework: if you want AWS billing and IAM AND full Anthropic API features, Claude Platform on AWS is the new right answer. If you want AWS billing and IAM AND a uniform Bedrock API across multiple model providers, Bedrock remains the choice. If procurement and IAM aren’t constraints, direct Anthropic API is fastest to set up and gets new features earliest.

For teams currently using direct Anthropic API and considering migration to Claude Platform on AWS, the lift is small. The API contract is identical; only authentication and the endpoint URL change. Migration is essentially a configuration swap, not a code rewrite. The bigger consideration is whether your organization’s billing and procurement preferences justify the move.

What’s next

Three threads to track over the next 90 days. First, region expansion. Anthropic launched in primary regions but enterprises with strict data-residency needs (specific EU countries, APAC, government regions) will push for broader region coverage. Watch for regional announcements as Anthropic and AWS scale the deployment.

Second, parity with Bedrock features that customers value. Bedrock has accumulated specific Bedrock-only features (Guardrails, Bedrock Agents, Knowledge Bases) that some customers depend on. Whether Claude Platform on AWS gains feature parity with these Bedrock additions, integrates with them, or stays as a clean Anthropic-API-only path will shape the long-term competitive dynamic between the two AWS-resident Claude options.

Third, the strategic implications for the Microsoft-OpenAI alignment. With Anthropic now offering a co-resident enterprise platform on AWS, the cloud-AI alignment story sharpens: Microsoft is the OpenAI cloud, AWS is increasingly the Anthropic cloud, Google Cloud is the Gemini cloud. Enterprises that want flexibility may opt for multi-cloud AI; enterprises that want simplicity will likely concentrate on whichever cloud-AI pair they already use most. Each of the three frontier labs now has at least one cloud-platform-grade distribution channel; the next round of competition is at the procurement and integration layer.

The strategic context for Anthropic specifically: with the company reportedly in negotiations on a $900B+ funding round, distribution depth matters as much as model quality for the next phase of growth. Claude Platform on AWS gives Anthropic enterprise reach it didn’t have through direct API alone. The combination of strong product, deep AWS distribution, and continued model improvements positions Anthropic to compete head-on with OpenAI for enterprise AI workloads through 2026 and 2027.

Frequently Asked Questions

How is Claude Platform on AWS different from Claude in Bedrock?

Claude in Bedrock is Bedrock’s hosted Claude — accessed via Bedrock’s API, with Bedrock’s feature set, and a curated subset of Anthropic’s capabilities. Claude Platform on AWS is Anthropic’s full platform, accessed via Anthropic’s native API, with AWS billing and IAM auth on top. The difference is API contract and feature surface: Bedrock = Bedrock API; Claude Platform on AWS = full Anthropic API.

Do I need a separate contract with Anthropic to use Claude Platform on AWS?

No. The AWS Marketplace subscription is sufficient for billing. Some enterprise customers may want a direct Anthropic relationship for SLAs, support escalation paths, or co-marketing — but the baseline procurement path runs entirely through AWS.

What regions is Claude Platform on AWS available in?

Anthropic launched in primary US and EU regions first. Specific region availability is in the AWS Marketplace listing; check it before planning deployment because the list expands over time. APAC and government regions are common follow-on launches for AWS Marketplace services.

How does cost compare across the three options?

Per-token pricing is similar across direct Anthropic, Bedrock, and Claude Platform on AWS. The economic differences come from procurement: AWS Enterprise Discount Programs apply to Claude Platform on AWS spend in the same way as other AWS services, which can produce meaningful effective discounts at scale. For organizations without AWS EDPs, the per-call costs are comparable across all three.

Can I migrate from Bedrock-mediated Claude to Claude Platform on AWS?

Yes, but it requires code changes. Bedrock’s API contract is different from Anthropic’s Messages API, so migration involves swapping request/response handling. The lift is moderate — typically a few days of engineering work for a non-trivial codebase. If you depend on Bedrock-specific features (Guardrails, Agents), evaluate whether equivalent functionality on Claude Platform on AWS meets your needs before migrating.

Does Claude Platform on AWS support Claude Managed Agents and Agent Skills?

Yes, both are part of the launch. This is one of the key differentiators from Bedrock-mediated Claude, where Anthropic-specific features like Managed Agents historically lag the direct API. Claude Platform on AWS gives enterprises access to those features with AWS billing and IAM auth out of the gate.

What about data residency and privacy?

Customer data flows terminate in AWS regions, with AWS networking and isolation primitives applied. Anthropic operates the platform infrastructure inside those regions. For enterprises with specific data-residency requirements (GDPR, sovereign cloud, regulated industries), this architecture is materially friendlier than calling api.anthropic.com over the public internet. Check the AWS Marketplace data-handling addendum for specific commitments.