cloudflare-workflows

Long-running workflows on Cloudflare Workers use the Step API (do, sleep, sleepUntil, waitForEvent) to orchestrate multi-step processes with automatic retries and state persistence. Steps execute durably across hours or days, with each step's state saved independently, enabling recovery from failures without re-executing completed steps.

Yes, durable workflows support configurable retry backoffs and timeouts per step. Each step can define its own retry strategy, allowing fine-grained control over how failures are handled across different stages of your workflow.

Use step.waitForEvent to pause workflow execution until external events arrive, enabling event-driven orchestration and human approval gates. State persists while waiting, and workflows resume automatically when events trigger, supporting multi-day approval cycles.

Workflow state persists per step with JSON-serializable values, so on failure or timeout, retries resume from the failed step without replaying completed steps. This isolation ensures idempotency and reduces wasted computation across retry attempts.

Yes, durable workflows return JSON-serializable values and deploy as Cloudflare Workers via Wrangler, integrating seamlessly into standard deployment pipelines while maintaining state persistence and retry guarantees across long-running tasks.

Use durable workflows for multi-step processes spanning hours to days, API orchestration requiring retries, scheduled tasks with state checkpoints, or event-driven workflows needing durability guarantees. Standard Workers suit stateless, request-response patterns.