Command reference · Cursor CLI
agent worker start
Starts the private worker and connects it to Cursor.
agent worker start starts the private worker process and connects it to Cursor, so cloud-agent tasks run in your own environment instead of Cursor's cloud. It's configured via flags like --worker-dir (workspace root, repeatable, the first value sets the assignment identity), --name, --label/--labels-file, --pool/--pool-name, --idle-release-timeout, and --management-addr for /healthz, /readyz, /metrics. --verbose enables verbose startup logs.
✅ WHEN TO USE IT?
Expose several repos at once as worker workspaces
agent worker start --worker-dir ~/repos/app-a --worker-dir ~/repos/app-b
Expose health/metrics endpoints for monitoring the worker
agent worker start --management-addr 0.0.0.0:9100
⛔ WHEN NOT TO?
You want to check up front whether auth and network would even work
agent worker start connects live immediately, without a dry-run check first.
Better: Run agent worker debug first, then agent worker start
Only a single, short agent run is needed
agent worker start starts a long-running process connected to Cursor.
Better: Use a normal session with agent or agent -p instead of running a worker
SOURCES
- Cursor Docs — CLI Reference: Parameters ↗ cursor.com