Run workflows

Execute local workflow files, resume failed runs, and inspect built-in workflow templates.

The CLI can execute workflow files directly, whether they are written in YAML, TypeScript, or Python.

Run a workflow file

agent-relay run relay.yaml

Supported file types:

  • .yaml and .yml
  • .ts and .tsx
  • .py

Validate before running

agent-relay run relay.yaml --dry-run

Dry runs validate the workflow and show the execution plan without starting the real run.

Resume or restart from a step

agent-relay run relay.yaml --resume <run-id>
agent-relay run relay.yaml --start-from implement --previous-run-id <run-id>

Use these options when a long workflow fails late and you want to retry without repeating earlier successful work.

List built-in templates

agent-relay workflows list

This prints the built-in YAML templates bundled with the CLI.

When to use cloud instead

Use agent-relay run when the workflow should execute locally against your current machine. Use agent-relay cloud run when you want the same workflow model in a sandboxed cloud environment.

See also