Schedules, webhooks, and events
Start work from time, an incoming request, or a completed workflow while keeping execution visible.
Choose a trigger
Automation creates a new Machine task or Arc run from a schedule, webhook, or event. A schedule uses an interval or cron expression. A webhook accepts an incoming request through its tokenized endpoint. An event trigger reacts to a Machine or Arc outcome.
Run the target manually first. Use a clear input template and verify the target's output, permissions, and timing before allowing repeated starts. Visitor demos begin with schedules disabled.
Overlap and trigger chains
A scheduled Machine can skip an occurrence when earlier work is still active, or queue another task when that overlap policy is selected. Arcs enforce their own active-run restriction. A skipped start is a recorded outcome you can inspect.
Event templates can include the source output and status. Trigger-chain depth is bounded to reduce accidental loops between targets. Prefer a workflow with a visible stopping point over a cycle that continually starts more work.
Durable delivery
Due schedules advance and enqueue their fire jobs in one transaction. Trigger job receipts tie a stable job identity to its outcome, so reclaiming the same job reuses its recorded result instead of creating another run.
Run creation, execution-job enqueueing, the trigger receipt, and schedule counters commit together. An infrastructure failure rolls that transaction back for retry. This protects the start of the workflow; individual external actions still need their own recovery and review boundaries.
The worker is part of the system
Submitting work and executing it are separate steps. The durable queue needs a worker to claim jobs. Hosted installations use authenticated worker requests from a trusted scheduler; a dedicated worker is available where a continuously running process is supported.
If tasks remain queued, first check worker activity and scheduler access. For live workloads, verify that the scheduler actually reaches the deployment and that a submitted task progresses through to its outcome.