OpenAI’s Agents Now Run on Amazon’s Cloud
OpenAI and Amazon launched Bedrock Managed Agents at DevDay 2026, enabling companies to run OpenAI agents entirely within AWS. Customer data remains secure and never leaves Amazon's cloud.
(September 29, 2026 - 6:08 pm)
Credit: OpenAI
Companies can now build OpenAI agents designed to operate seamlessly within Amazon Web Services (AWS), keeping customer data within AWS. OpenAI integrated Bedrock Managed Agents, powered by its own agent framework, directly into its DevDay 2026 event in San Francisco on Tuesday. Although currently in limited preview, the AWS product page offers more details.
OpenAI developed this solution with Amazon, tailoring it to function natively within AWS and integrate with AWS resources. As outlined in their DevDay recap, the agents have unique identities and meticulously log every action for auditing purposes.
All inference runs on Amazon Bedrock, ensuring customer data remains secure within AWS. AgentCore, Amazon’s platform for building and running agents, serves as the default compute environment. Future enhancements to Bedrock Managed Agents include authorization policy enforcement, agent discovery, observability, and evaluation tools.
Salesforce, a notable early adopter, leverages OpenAI models and AWS infrastructure alongside Salesforce data and controls.
“Data never leaves AWS, and security stops being the bottleneck between our customers and what comes next,” expressed Joe Inzerillo, Salesforce’s president of enterprise and AI technology, in a statement on the AWS page.
Previously, in August, Salesforce placed Anthropic’s Claude at the heart of its products. Notably, AWS reported that nearly 90% of their early agent prototypes never progressed to shipping.
Key Features of Bedrock Managed Agents:
- Built on OpenAI's own agent framework.
- Handles inference, memory, and skills within the customer’s environment.
- Secure: Customer data remains within AWS.
- Agents have unique identities and log every action.
- Supports computer use through API access and tools from Codex.
Recent OpenAI Developments:
On Monday, OpenAI disclosed that its agents had inadvertently posted users' images online, affecting 53 ChatGPT users. The same event also saw the launch of always-on dot agents, plus a new $500 Pro plan and halved usage limits for Pro 200 plans.