Skip to main content

Options

The Agent Prompt supports single-select and multi-select modes, a free-text option, and multi-step flows.

Selection Modes

Each mode drives a different selection interaction. Single-select uses radio semantics; multi-select uses checkbox semantics.

Single Select

Each option row acts as a radio — selecting one deselects all others. The selected row is outlined to confirm the choice.

Multi Select

Each row carries a checkbox, and any number of options can be selected. The resolved answer is the set of checked values.

Free Text

The “Something else…” free-text row renders when onOtherChange is wired. In single-select, the row is mutually exclusive with the option rows — typing deselects any chosen option, and selecting a row retains but subordinates the typed text. In multi-select, the typed value is an independent addition to the selection.

Multi-step

Passing a step object switches the action row to a Back/Next/Send stepper and shows the current step count in the header. The first step shows only Next; middle steps show both Back and Next; the last step shows Back and Send.

Behavior

The primary action is always enabled — attempting to send with nothing selected surfaces an inline error rather than disabling the button.

Visual States

The prompt renders in an empty (no selection), selected, and disabled state. The disabled state dims the options and action row but keeps the close button active so the user can always dismiss.

Location

The Agent Prompt occupies the same position as the Prompt Bar — replacing it for the duration of the question. Once the user sends an answer, the Prompt Bar returns.

Empty Answer Error

When the user presses Send or Next without selecting an option or entering free text, an error message appears beneath the options. Selecting any option or typing in the free-text field clears the error immediately.

Keyboard Interaction

Users can navigate and operate the Agent Prompt using standard keyboard controls.
Last modified on July 24, 2026