AWS Step Functions integrates with Amazon Bedrock AgentCore for AI reasoning
AWS Step Functions now supports AI agent reasoning steps via integration with Amazon Bedrock AgentCore, currently in preview. This feature automates complex tasks like document classification or data extraction within workflows, allowing for parallel execution, human approval, and detailed audit trails. It's available in select AWS Regions, with standard pricing for both services applying.
- →Integrate AI agent reasoning into AWS Step Functions workflows
- →Enhanced visibility and auditability for agentic reasoning steps
- →Flexible agent configuration and reuse
- →Agent context persistence and regional availability
Features (1) ›
- Integrate AI agent reasoning into AWS Step Functions workflows
AWS Step Functions now allows the addition of AI agent reasoning steps through an integration with Amazon Bedrock AgentCore, which is currently in preview. This enables automation of tasks like document classification and data extraction within workflows.
Enhancements (2) ›
- Enhanced visibility and auditability for agentic reasoning steps
Workflow execution history provides detailed insights into agent input, output, token usage, and duration, with links to CloudWatch for tracing agent decisions. This aids in auditing and understanding agent behavior.
- Flexible agent configuration and reuse
Agents can be defined using an AgentCore harness, which specifies the model, tools, and behavior, and can be reused or created within Workflow Studio. Per-invocation overrides allow context-specific adaptation without configuration duplication.
Notes (1) ›
- Agent context persistence and regional availability
Agent context can be persisted across workflow invocations using a session ID. The harness integration is available in preview in specific AWS Regions: US East (N. Virginia), US West (Oregon), Europe (Frankfurt), and Asia Pacific (Sydney).
https://aws.amazon.com/about-aws/whats-new/2026/06/aws-step-functions-agentcore/