Auto mode

Interactively, Syntic Code pauses whenever it wants to do something you have not pre-approved, showing a prompt so you can accept or reject the action. That safety check is exactly what you want at a keyboard, but it stops the agent dead in a pipeline where no human is watching. Auto mode runs syntic without interactive prompts so it can complete a task from start to finish on its own, which makes it the foundation for CI jobs, batch refactors, and scheduled automation.

What auto mode changes

When you run in auto mode, the agent stops asking for confirmation and instead decides each action against the permission rules already in your settings. Anything in the allow list proceeds; anything in the deny list is refused; and rather than prompting for the ambiguous middle, the agent follows the non-interactive policy you have configured. Because there is no human backstop, your permission rules are the entire safety boundary, so they must be written deliberately before you turn auto mode on.

Running it

Invoke the CLI non-interactively, passing the task as an argument and selecting auto mode so it never waits for input:

syntic --auto "update the changelog for the pending release and open a pull request"

In a pipeline, capture the exit code and output so the job succeeds or fails visibly. Provide credentials through environment variables from your secret manager, and pin the model tier in settings so cost and behavior are predictable across runs.

Using it safely

Auto mode should run with the tightest permissions that still let the job finish. Deny destructive shell commands, scope filesystem access to the working tree, and prefer running inside a sandbox or an ephemeral container so a mistake cannot reach the wider system. Log the full transcript of each run for review, and start new automation in a throwaway branch or a dry-run configuration until you trust the results.