
context.step()) for automatic retries and checkpointing, and Waits (context.wait()) for suspending execution without compute charges. Additional operations include waitForCallback() for human-in-the-loop approvals, waitForCondition() for polling, and parallel() or map() for concurrency patterns.AWS Compute Blog
A deeper technical walkthrough covering the checkpoint and replay mechanism, how durable operations are persisted, and a user onboarding example with email verification and 24-hour timeout using waitForCallback.
Architecture guidance
Durable functions focus on application logic orchestration using familiar programming languages. Step Functions are suited for workflow orchestration across AWS services with visual state machines. Choose based on whether your coordination is code-centric or service-centric.
re:Invent 2025
Eric Johnson and Michael Gasch walk through the evolution from monolithic architectures to durable functions, covering the checkpoint and replay mechanism, compensation patterns, the testing framework, and pricing model.
What to build
Durable functions are well suited for multi-step workflows that need reliability and long-running coordination.
Developer resources
The SDK provides the primitives for checkpointing, retries, and execution flow management. Available for Python and Node.js/TypeScript.
re:Invent 2025
This session covers three major launches: the Serverless MCP server, Lambda Managed Instances, and Lambda Durable Functions, with a live-built application demonstrating each capability.
Build multi-step applications and AI workflows with AWS Lambda durable functions
AWS News Blog
The launch announcement walks through creating a durable function, using the SDK primitives, and building a production-ready order processing workflow with callbacks for external approvals and configurable retry strategies.