Troubleshooting
What the common failures mean.
Invalid User API Key
The key is missing, malformed, or was revoked. Check that the secret is set on the repository (or organisation) and exposed to the job.
[plan_required] ... not available for free users
The key is valid but the account cannot run agents. This is billing, not configuration: nothing in the workflow will fix it.
A model id is rejected
When Cursor rejects the model, the step log and the job summary list every model and setting the API key can use, written the way the model input takes them. Copy one of those lines.
Cannot use this model: <id>means the id is not in the account's catalog. The catalog no longer hasdefault, so usecomposer-2.5, which is also the input's default. The Cursor CLI's names such asgrok-4.7-mediumare not catalog ids either. Writegrok-4.7orgrok-4.7[reasoning_effort=medium]instead.Model unavailable on Plus ... with fixed settingsmeans the plan does not allow a setting you overrode. Remove the brackets to run the catalog's defaults.Invalid parameters for registry modelmeans a setting id or value does not exist for that model. Check the spelling against the list.Invalid 'model' valuemeans the brackets are malformed. The action stops before it starts the agent.
The step timed out or failed
Check the job summary. When a run reports cancelled (including timeout-triggered cancellation) or error, the action exits with code 1. When timeout elapses the action asks the SDK to cancel the run and waits up to ten seconds for it to stop; if it does not, the step still fails as cancelled. Increase timeout if the agent needs more time to complete.
Cannot find module
The action failed to install its own dependencies, which points at the runner's network or a corrupted npm cache rather than at your workflow. Re-run the job; if it persists, open an issue with the full step log.
Last updated on