Modules · Meeton Chat · Step 2
Prompt
The foundation this agent starts from: the platform default, or your own override. Most teams never rewrite it; they add a few lines of instructions on top and let the default do the rest.
Base prompt
Special tokens
The agent does not call your forms directly. It emits a token inside the text of its answer, and the widget replaces that token with the right form or action. This is why answers are rendered whole rather than streamed: the widget needs the complete message to know what to draw. The tokens an agent may use are listed on the Prompt step as magic variables, and each agent type has its own set.
| Token | What the widget draws | Agent |
|---|---|---|
[USER_FORM] | The lead form, with the contact fields configured under Lead form. | AI SDR |
[CALENDAR_FORM] | The booking popup for the calendar the routing rules pick. | AI SDR |
[MEETINGS_FORM] | The visitor's own meetings, to reschedule or cancel ("manage my meetings"). | AI SDR |
[INQUIRY_FORM] | The inquiry form, for a question the agent should hand to a person. | Both |
[SUPPORT_REQUEST_FORM] | The support request form. | Customer support |
[CONFIRM_RESOLUTION] | "Did that solve it?" buttons. | Customer support |
[CSAT_FORM] · [FEEDBACK_FORM] | The satisfaction rating and feedback form after a resolution. | Customer support |
Document recommendations work the same way: the agent references Library documents by ID, and the widget renders them as cards under the message, independently of any form token.
Context data packs
The prompt is assembled per message from the base prompt, your instructions, the knowledge that matches the question, the tools the agent may use, and the visitor's context: pages viewed, referrer, documents read, earlier messages, contact status, recognised company and target-account match. The Prompt step shows which packs are in play so you can see what the agent knows before it answers.