What is AWS Step Functions?
Learn how Step Functions uses state machines, tasks, service integrations, workflow types, retries, callbacks, and execution history.
What is AWS Step Functions?
AWS Step Functions is a managed workflow service for building distributed applications, automating processes, orchestrating microservices, and coordinating data or machine-learning pipelines. A workflow is a state machine whose states define tasks and control flow.
How this topic fits
Workflows
Coordinate multi-step work with visible state, retries, timeouts, compensation, approvals, and long-running business processes.
State machines, states, and executions
A Step Functions workflow is defined in Amazon States Language. Each state performs work or controls the flow by choosing, waiting, running branches in parallel, iterating over items, succeeding, or failing.
Every run is an execution with its own input, state transitions, and outcome. That explicit history makes the control path visible and separates workflow state from the runtime of any individual Lambda function or service task.
Call Lambda, supported AWS service integrations, activities, or approved external endpoints to perform work.
Branch, wait, map, run in parallel, pass data, or end the execution with an explicit outcome.
Shows state transitions and outcomes so developers and operators can inspect the workflow path.
Choose the workflow type from the workload
Step Functions provides Standard and Express workflows with different duration, execution, delivery, history, and pricing characteristics. Do not choose only from expected volume; auditability, integration pattern, duplicate tolerance, and maximum execution time also matter.
Keep the choice documented with the business operation. A workflow type that fits a high-rate transformation may not fit a long-running approval process, even if both use the same task services.
Service integrations and callbacks
Step Functions can call AWS services directly through optimized or AWS SDK integrations, which can remove functions that only translate one API request into another. The workflow role must still receive the narrow permissions required for those calls.
Request-response tasks continue after an API response. Other supported patterns can wait for a job or a callback token. Callbacks are useful for human or external work, but require timeout, authentication, duplicate, cancellation, and recovery decisions.
Failure handling and data flow
Configure retries for known transient errors with bounded backoff, and use catches for errors that need an alternate path. Avoid retrying validation or authorization failures that cannot succeed without a change in input or policy.
Transform and pass only the data each state needs. Large or sensitive payloads can increase cost, exposure, and coupling; store durable documents in an appropriate data service and pass stable references through the workflow.
Go deeper with AWS
Read the primary AWS documentation behind the service definitions and architecture guidance in this guide.
Next steps
Continue with the concepts that most directly shape this decision.
What are serverless workflows?
Learn when explicit workflow state improves retries, branching, parallel work, human approval, visibility, and recovery in distributed applications.
What is Amazon EventBridge?
Understand EventBridge event buses, rules, pipes, schemas, targets, schedules, permissions, delivery, and recovery in event-driven AWS systems.
Retries, DLQs, and destinations
Design finite retry policies, classify failures, isolate poison work, understand invocation-specific behavior, and build a safe redrive process.
Serverless generative AI on AWS
Combine Amazon Bedrock with APIs, functions, workflows, events, data, guardrails, evaluation, observability, and bounded tool execution.
Questions about this topic
Browse all serverless concepts
Explore compute, APIs, events, workflows, data, security, reliability, operations, delivery, and AI from one connected AWS serverless learning path.
What is serverless?
Learn what serverless computing means, how it works on AWS, what the cloud provider manages, and which trade-offs still belong to your team.
What is AWS Lambda?
Understand how AWS Lambda runs code, how functions are invoked, where permissions and state live, and when Lambda is a good compute choice.
What is serverless architecture?
Learn how managed compute, APIs, events, queues, workflows, data, identity, and observability fit together in a serverless architecture on AWS.
AWS serverless security
A practical guide to identity, least privilege, event validation, secrets, data protection, dependencies, and incident visibility in AWS serverless systems.
What is agentic AI?
Learn what makes an AI system agentic, how models, tools, memory, orchestration, identity, and evaluation fit together, and where serverless patterns help on AWS.
What is cloud computing?
Understand how on-demand infrastructure and managed services change the way teams provision, scale, pay for, and operate software systems.
What is event-driven architecture?
Learn how producers, routers, and consumers communicate through events, and how to design schemas, delivery, retries, and ownership deliberately.
What is a serverless API?
Understand the request path, authentication, routing, compute, data, throttling, observability, and failure controls behind a serverless API on AWS.
What are serverless workflows?
Learn when explicit workflow state improves retries, branching, parallel work, human approval, visibility, and recovery in distributed applications.
What is Amazon API Gateway?
Learn how API Gateway creates, publishes, secures, monitors, and operates HTTP, REST, and WebSocket APIs for AWS and external backends.
What is Amazon EventBridge?
Understand EventBridge event buses, rules, pipes, schemas, targets, schedules, permissions, delivery, and recovery in event-driven AWS systems.
Make the next serverless decision explicit.
Explore focused products for repeatable work, or bring the architecture decision that needs context and experienced review.