OpenAI models, Codex, and Managed Agents come to AWS
Key Points
- GPT-5.5 and OpenAI models available on Amazon Bedrock (limited preview)
- Codex can run on Bedrock with CLI, desktop, and VS Code integrations
- Bedrock Managed Agents enable production-ready agents with AWS security
Summary
OpenAI and AWS expanded their partnership to bring OpenAI frontier models (including GPT-5.5), Codex, and Amazon Bedrock Managed Agents to AWS. All three offerings are launching in limited preview and run inside Amazon Bedrock so enterprise customers can use these capabilities within existing AWS security, identity, billing, and compliance frameworks.
Key Points
- OpenAI models on Amazon Bedrock: GPT-5.5 and other frontier models are available via Bedrock, enabling AI apps, embedded intelligence, and agentic workflows inside AWS environments.
- Codex on Bedrock: Codex (coding agent and product suite) can be configured to use Amazon Bedrock as the provider via the Bedrock API. Integrations include the Codex CLI, Codex desktop app, and the Visual Studio Code extension.
- Data and billing: All customer data is processed by Amazon Bedrock; eligible customers can apply Codex usage toward AWS cloud commit billing.
- Bedrock Managed Agents (powered by OpenAI): Managed Agents support context retention, multi-step workflows, tool use, orchestration, and enterprise-grade governance and security, streamlining path to production.
- Availability & requirements: Features are limited preview—Bedrock access and, for some benefits, an AWS commitment are required. Expect enterprise integration work around IAM, logging, and compliance.
Practical actions for engineers
- Request access to the limited preview and confirm Bedrock availability in your account.
- Review IAM, logging, and compliance controls for Bedrock integrations before routing data.
- Configure Codex to use Bedrock via the Bedrock API and validate integration with Codex CLI / desktop / VS Code.
- Evaluate billing treatment if you have applicable AWS commit credits and update procurement workflows accordingly.