← Platform
Durable execution
Runs for minutes or months. Survives every failure.
Durable event history. Crashes, network hiccups, waiting periods — the run picks up exactly where it paused.
Run
invoice-factoring-wf-2847
Running
14:22:01 workflow.start
Completed
run_id: wf-2847 · —
14:22:03 read_invoice_package
Completed
credit reasoning model · 1.8s
14:22:05 validate_amount
Retried
2 retries · network · 4.2s
14:22:09 request_human_approval
Waiting on human
conf 0.62 · $28,400 over $25K policy · 2h 14m
— policy_check
Pending
pending
— workflow.complete
Pending
pending
Event-sourced history Replayable · durable
Run in progress · retried twice · waiting on human signal
Guarantees
The primitive your runtime couldn't give you.
Exactly-once execution
Side effects run once even across crashes.
Details
Deterministic replay
Same inputs, same outputs. Always.
Details
Long-running state
Workflows span weeks without holding memory.
Details
Infrastructure-aware
Built for retries, failover, and recovery.
Details
What durability gets you
Workflows that outlive the servers running them.
Replay any run, exactly
Every step written to history. Debug by replaying bit-for-bit.
Details
Human signals, days or weeks
Workflows wait on signals without consuming resources.
Details
Backoff, jitter, compensation
Retry policies are declarative. Saga patterns without bespoke plumbing.
Details
Durability you can bet production on.
Scope one workflow and see it run.