Running Workflows
Once you've set up a project with a team, agents, and a workflow, you're ready to execute it.
Manual Execution
Open a workflow and click Run. This creates a new Workflow Run and starts execution immediately.
The run page shows:
- Overall status — pending, running, completed, stopped, or error.
- Stage progress — which stage is currently executing and which have completed.
- Stage results — the output from each stage as it completes.
- Timing — when the run started, how long each stage took.
Scheduled Execution
Use Triggers to run workflows automatically:
- Open a workflow and create a new trigger.
- Define a schedule — natural language ("every Monday at 9am EST") or cron expression.
- Set a timezone and optional start/end dates.
- The workflow runs automatically on schedule.
Use Fire Now to trigger a scheduled workflow immediately without waiting for the next scheduled time.
External Triggers
Workflows can be triggered programmatically via the API:
POST /api/v1/trigger
This lets external systems — CI/CD pipelines, monitoring alerts, other applications — kick off workflows on demand. See API for details.
Monitoring Runs
Browse all runs for a project from the project page, or filter by workflow. Each run is a complete record of the execution — status, timing, and per-stage results.
Runs in an error state show which stage failed and the error details, helping you diagnose and fix issues.
Stopping a Run
Running workflows can be stopped from the run detail page. This halts execution at the current stage.
Re-running from a Stage
After a run has completed, stopped, or errored, you can start a new run from any stage in the original run's timeline. Hover over a stage divider in the event timeline to reveal the Re-run from here button.
The new run:
- Skips all stages before the selected starting point.
- Inherits shared memory -- stage results produced by earlier stages in the source run are injected into the new run, so downstream agents have the same context they would if those stages had actually executed.
- Inherits task summaries -- agents receive summaries of what happened in prior stages, maintaining continuity.
- Continues through remaining stages to the end of the workflow, following the normal stage order (including outcome routing).
Each re-run is a fully independent workflow run with its own events and results. The source run is never modified.
When to use it
| Scenario | Benefit |
|---|---|
| A late stage errored due to a tool misconfiguration | Fix the config and re-run from the failed stage instead of re-executing earlier stages |
| You want to tweak a stage's prompt and see the effect | Re-run from that stage without waiting for upstream stages to repeat |
| A run completed but you want to try a different outcome path | Re-run from the branching stage to explore the alternative route |