How Amazon Developers Can Now Use Claude Code and Codex for Agentic Coding

Introduction

Amazon recently made a significant shift by approving third-party agentic coding tools for its developers. Following an internal rebellion where employees pushed back against restrictions on tools like Claude Code, the company has now authorized Anthropic's Claude Code and OpenAI's Codex for production use. These tools run on Amazon Web Services (AWS) and Amazon Bedrock, giving developers secure, scalable access without complex setup. This guide walks you through how to get started with these tools, from understanding the policy change to integrating them into your workflow.

What You Need

• Amazon developer account – with access to internal development resources
• AWS account – with permissions to use Amazon Bedrock
• Access to Amazon Bedrock – the fully managed service for generative AI applications
• Understanding of agentic coding tools – familiarity with AI-assisted coding concepts
• IDE or CLI environment – such as Visual Studio Code, IntelliJ, or terminal

Step-by-Step Guide

Step 1: Understand the Policy Change

As of May 12, Amazon has officially approved Claude Code and Codex for production use. Previously, internal communications mandated using Amazon's own Kiro tool for production code and required special clearance for third-party tools. Now, developers can use Claude Code and Codex without any special approval. This change came after employee feedback, including a discussion thread with about 1,500 endorsements urging the company to adopt Claude Code. Knowing this context helps you confidently adopt the tools.

Step 2: Access Claude Code and Codex via Amazon Bedrock

Both tools are available through Amazon Bedrock, which simplifies infrastructure management and ensures data security. To access:

1. Log in to the AWS Management Console and navigate to Amazon Bedrock.
2. In the Bedrock dashboard, look for Foundation Models or Agentic Tools section.
3. Select Claude Code (Anthropic) or Codex (OpenAI). Claude Code is available immediately; Codex will follow soon.
4. Enable the model in your region. Amazon runs these tools on AWS infrastructure, so you don't need to set up separate servers or manage capacity.

Step 3: Configure Your Environment

Once the tools are enabled in Bedrock, configure your development environment:

1. Set up API keys or IAM roles to authenticate with Bedrock. Follow AWS security best practices.
2. Install any required SDKs or plugins. For example, Claude Code may have a CLI tool or IDE extension.
3. Test connectivity by sending a simple prompt (e.g., "Write a Python function to sort a list").
4. Ensure your network allows outbound connections to AWS endpoints for Bedrock.

Step 4: Integrate Tools into Your Workflow

Use Claude Code or Codex directly in your coding environment:

• In IDE: Use plugins that connect to Bedrock. These tools can suggest code, generate functions, or debug issues.
• In CLI: Run commands like claude-code "generate a React component" to get instant outputs.
• In CI/CD pipelines: Automate code reviews or generate unit tests using the API.

Amazon VP Jim Haughwout stated: "To help you invent more for customers, we are expanding the agentic AI tool available to you." Use these tools to accelerate development, but always review outputs carefully.

Step 5: Review and Validate AI-Generated Code

As Satyam Dhar, staff software engineer at Galileo, notes: "Constraint now needs to shift to review, validation, and making sure the system behaves the way you think it does." Follow these practices:

1. Always review generated code for logic errors, security vulnerabilities, and adherence to Amazon's coding standards.
2. Run automated tests (unit, integration, and security scans) on AI-generated code.
3. Use version control to track changes. Tag commits that include AI assistance for easier auditing.
4. Document any modifications you make to the tool's output.

Step 6: Keep Using Kiro as an Alternative

Amazon's own Kiro agentic coding service (built on Bedrock) remains available. As an Amazon spokesperson said: "Our builders are using Kiro for agentic coding, and now with both Claude Code and Codex running on AWS, we are making additional tools available as well." You can choose the best tool for each task. For example, use Kiro for projects requiring deep integration with Amazon-specific services, and Claude Code or Codex for general-purpose coding assistance.

Tips for Success

• Manage capacity: Because tools run on AWS, you can scale usage up or down without manual provisioning. Monitor usage via AWS CloudWatch to avoid unexpected costs.
• Focus on security: All data stays within AWS; you control access via IAM. Never share credentials or expose API keys.
• Encourage team adoption: Share best practices in code reviews. The internal response showed strong desire for these tools – leverage peer knowledge.
• Stay updated: Both Anthropic and OpenAI release model updates. Check Amazon Bedrock announcements for new versions or features.
• Reorder your evaluation metrics: As Dhar suggests, leadership may need to rethink productivity measurement. Focus on outcomes like code quality and time saved, not just lines written.

By following these steps, you can seamlessly integrate Claude Code and Codex into your Amazon development workflow, benefiting from top-tier AI assistance while maintaining security and compliance.