Agent modes

Every chat runs in one of three modes. Switch with the mode button under the message box, with ⇧Tab, or by typing /plan, /ask or /auto.

Mode	What happens
Plan	Hex reads and researches, then proposes a step-by-step plan. Nothing is changed until you approve it. After you approve, it works through the plan and ticks off steps.
Ask	Hex reads freely but asks before every edit and every command.
Auto	Hex edits and runs commands on its own, within your permission settings. Destructive commands still ask.

Set the mode new chats start in under Customize → Agent → Behavior. A project can also set its own default mode.

Approving actions

When Hex needs your OK, the step shows Allow, Always allow and Deny. Always allow remembers the choice for that tool in this project's folder. You can review and remove those choices in Customize → Agent → Permissions → Project permissions.

Thinking

Reasoning models can think before answering. Use the Thinking button (or /effort) to choose Off, Low, Medium or High. Higher takes longer and uses more of the model's context.