Skip to main content

​
What Are Starting Actions?

Starting Actions are pre-configured prompt chips that appear in the chat UI before the user sends their first message. They sit below the agent’s greeting and let users start common tasks with one click instead of typing. Admins define each action with a name, icon, prompt template, and optional variables.

​
Where They Appear

Starting Actions render as clickable chips in the agent chat interface, directly under the agent’s opening message. After the user sends their first message, the chips are no longer shown for that conversation.

​
Create Starting Actions

  1. In Intelligence, select the agent.
  2. Click Configure.
  3. Click + Add Starting Action in the Starting Actions section.
  4. Add an icon, name, and prompt template. Use {{variableName}} placeholders for any inputs you want to collect from users.
  5. Configure each detected variable. In the Detected Variables section, click a variable chip to open its Configure Variable dialog. See Configure Detected Variables below for the available settings.
  6. The action is automatically enabled. Toggle it off if you want to wait before enabling it for users.
  7. Click Add Starting Action.
Per-action settings:
  • Name – Label shown on the chip
  • Icon – Icon for the chip
  • Prompt template – The text sent when the user clicks the action; can include {{variableName}} placeholders
  • Enabled – Toggle to show or hide the action
  • Display order – Order of chips; reorder via drag-and-drop
Actions can be reordered with drag-and-drop and individually enabled or disabled.

​
Configure Detected Variables

Every {{variableName}} placeholder in the prompt template is auto-detected and listed under Detected Variables. Each variable must be configured before the Starting Action can collect input from users. Click a variable chip to open its Configure Variable dialog, set the following, then click Save:
  • Display Label – The label shown to users when filling in the variable.
  • Placeholder Text – Hint text shown inside the input field.
  • Input Type – How users enter the value: Text, Number, or Select from Options (dropdown).
  • Required – Toggle on to require a value before the form can be submitted.
  • Options (Select from Options only) – Add each dropdown choice as a value/label pair. Options are static and entered manually.
Variable settings — particularly Select from Options lists — must be updated manually. They are not populated dynamically from Element field values or other underlying data. When the source data changes (for example, when new records are added to a referenced Element), edit the Starting Action and update the affected variable’s options to match.

​
Variable Support

Use {{variableName}} in the prompt template to collect input before sending.
  • Placeholders are auto-detected from the template and can be configured individually.
  • Variable types — Text, Number, and Dropdown are supported.
  • Validation — Optional rules such as pattern, min/max length, and custom error messages.
Actions without variables send the prompt immediately when clicked. Actions with variables display an inline form to collect values before sending.

​
Runtime Behavior

Action typeBehavior
No variablesClicking the chip sends the prompt immediately.
With variablesClicking the chip expands an inline form; after the user fills it, the prompt is sent with the values substituted.
Chips disappear after the user sends their first message (whether from a Starting Action or by typing).
  • Building Agents – Agent creation, configuration, and chat experience, including Starting Actions in the UI