Claude Platform on AWS: Full Native API, Same-Day Features, IAM Auth

Anthropic has made the Claude Platform generally available on AWS. This isn't just another model hosting option—it's the full Claude API, including features that previously required direct Anthropic access, now available through AWS IAM, CloudTrail, and consolidated billing.

Starting today, AWS customers can deploy agents at scale with Claude Managed Agents (beta), build with code execution, skills, and the advisor strategy. All new features ship the same day they go live on the native Claude API.

What's Included: The Full Feature List

The platform includes:

  • Claude Managed Agents (beta) — build and deploy agents at scale
  • Advisor strategy (beta) — agents consult an advisor model for better decisions
  • Web search and web fetch — real-time data from the web
  • Code execution — run Python, create visualizations, analyze data within API calls
  • Files API (beta) — upload and reference documents across conversations
  • Skills (beta) — teach Claude best practices for consistent results
  • MCP connector (beta) — connect to any remote MCP server without writing client code
  • Prompt caching — reduce costs and latency on repeated context
  • Citations — ground responses in source documents
  • Batch processing — high-volume asynchronous workloads

Also included: Claude Console (prompt improver, generator, evaluation tools) and models Claude Opus 4.7, Sonnet 4.6, and Haiku 4.5.

How It Works: AWS IAM Auth, CloudTrail, Single Bill

Authentication runs through AWS IAM. Audit logging goes to CloudTrail. Billing appears on a single AWS invoice that fully retires against existing commitments. Teams use their existing AWS credentials and IAM policies—no new permissions to learn.

Claude Platform on AWS will be available in most AWS commercial regions and support global and U.S. inference geographies.

Customer Quotes: Real Deployments

Jonathan Echavarria, Principal Research Scientist (customer): "Claude Platform on AWS helped simplify how we access Claude, improved the experience for key users like our Claude Code engineers, and gave us a practical path to integrate further frontier AI capabilities into our cybersecurity and engineering workflows, while staying within our existing cloud operating model."

Tomas Oliva, AI Platform Engineer at OpenRouter: "Using Claude Platform on AWS gives OpenRouter and our users direct access to the latest and greatest features of the native Claude API... It has delivered consistent performance on uptime, latency, and throughput."

Avinash Vishwakarma, Chief Architect (customer): "Support has been one of the best parts, and it felt like one team across both companies, not two separate relationships."

Claude Platform on AWS vs. Claude on Amazon Bedrock

Both enable building on Claude models. The difference is who operates the service and feature availability.

Claude Platform on AWS — Anthropic operates the service. Data is processed outside the AWS boundary. You get all native Claude API features from day one. Good for companies wanting the full Claude Platform experience.

Claude on Amazon Bedrock — AWS is the data processor. Data stays within the AWS boundary. Good for strict regional data residency requirements or when data must be processed exclusively inside AWS.

Getting Started

The Claude Platform on AWS is available today. Visit the Claude Platform on AWS or explore the documentation.

If you have an existing Bedrock private offer, contact your Anthropic or AWS account executive before starting—discounts cannot be applied retroactively.

Technical Deep Dive: Code Execution and MCP Connector

Two features stand out for developers building agentic workflows:

Code execution lets you run Python code directly within API calls. Instead of spinning up a separate Lambda or EC2 instance for data analysis, you can pass a prompt like:

{
  "model": "claude-opus-4-7",
  "tools": [{"type": "code_execution"}],
  "messages": [{"role": "user", "content": "Generate a scatter plot of the Iris dataset and return the PNG"}]
}

Claude runs the code serverlessly and returns the output.

MCP connector (beta) allows Claude to connect to any remote MCP server without you writing client code. This means you can plug Claude into your existing data sources, databases, or internal tools with minimal glue code.

Why This Matters for AWS Developers

If you're already on AWS, this removes the friction of managing separate API keys, billing, and IAM policies for Claude. You get day-one access to new features, consistent latency, and support that feels like one team across both companies.

For teams building agentic systems, the combination of Managed Agents, code execution, and MCP connector means you can deploy production agents without standing up separate orchestration infrastructure.

What to Do Now

  1. Go to the Claude Platform on AWS page and enable the service in your AWS account.
  2. If you have existing Bedrock private offers, contact your account exec before using the new platform.
  3. Try the code execution tool with a sample dataset.
  4. Set up a Managed Agent for a simple internal workflow (e.g., summarizing support tickets).
  5. Check the documentation for the MCP connector to connect Claude to your existing services.