OpenAI and AWS Partner on Stateful Runtime for AI Agents in Bedrock
OpenAI and Amazon have announced a partnership to deliver a Stateful Runtime Environment that runs natively in Amazon Bedrock. The new runtime is designed to make multi-step AI workflows more reliable by handling persistent orchestration, memory, and governance automatically.
The Problem With Stateless Agent APIs
Most AI agent prototypes built on stateless APIs handle simple use cases: one prompt, one answer, perhaps one tool call. Production workflows are different. Real work unfolds across many steps, requires context from previous actions, depends on multiple tool outputs, and needs approvals and audit trails.
Stateless APIs force development teams to build their own orchestration layers. Teams must figure out how state is stored, how tools are invoked, how errors are handled, and how long-running tasks resume safely after interruptions.
What the Stateful Runtime Environment Provides
The Stateful Runtime Environment, jointly designed by OpenAI and AWS, runs inside your AWS environment and is optimized to work with AWS services. Instead of manually stitching together disconnected requests, agents automatically execute complex steps with working context that persists.
This working context includes memory and conversation history, tool and workflow state, environment usage patterns, and identity and permission boundaries. Teams can focus on workflow logic and business requirements rather than scaffolding infrastructure.
Use Cases Enabled by Stateful Runtime
The runtime makes it easier to deliver production solutions that were previously difficult to build with stateless APIs. Common use cases include multi-system customer support workflows, sales operations with handoffs, internal IT automation, and finance processes requiring approvals and audits.
When the runtime handles persistent orchestration and state across steps, agents can reliably execute tasks that span minutes, hours, or even days without losing context.
AWS-Native Security and Governance
The runtime is designed to operate within customers existing AWS environments, making it easier to comply with current security postures, tooling integrations, and governance rules. Stateful tasks carry forward the context and control boundaries needed for multi-step execution in regulated environments.
AWS customers will have access to the Runtime Environment powered by OpenAI models, optimized for AWS infrastructure and tailored specifically for agentic workflows.
Availability and Getting Started
The Stateful Runtime in Amazon Bedrock will be available soon. Organizations interested in exploring the platform can reach out to their OpenAI team or request contact through OpenAI or AWS channels.
This partnership represents a significant step toward making AI agents production-ready for enterprise workloads, reducing the operational burden on development teams while maintaining security and compliance standards.
FAQ
What is a stateful runtime for AI agents?
A stateful runtime maintains context, memory, and workflow state across multiple steps of an AI agent execution. Unlike stateless APIs that treat each request independently, a stateful runtime preserves working context including history, tool outputs, and permission boundaries.
How does this differ from using OpenAI APIs directly?
The OpenAI API is stateless by design. Each request is independent. The Stateful Runtime Environment handles persistence, orchestration, error recovery, and state management automatically, allowing agents to execute multi-step workflows reliably.
Do I need to migrate existing AWS infrastructure?
No. The runtime runs natively within your existing AWS environment and is optimized to work with AWS services you already use. It is designed to comply with your current security posture and governance rules.
When will the Stateful Runtime be available?
OpenAI and AWS have announced the partnership and planned availability. The runtime will be available soon, with interested customers able to request early access through their OpenAI or AWS contacts.
