Playground
Talk to your agent the way your users will — the real widget, your active design, live answers — before you embed it.
The Playground (/playground) mounts the real chat widget in a live conversation against your active environment’s catalog. It is the fastest way to see how the assistant answers before you drop the widget into your own app.
What the playground does
Every message is a real agent turn: the assistant reasons over your uploaded catalog and knowledge exactly as it will in production. It follows the dashboard’s active environment, so switching between development and production starts a fresh conversation against that environment’s catalog.
The controls
A control bar above the stage lets you preview the widget in different states without leaving the page:
- Design — preview any saved widget design, or the workspace default.
- Size — Compact, Comfortable, or Expanded.
- Layout — normal or split.
- Scheme — light or dark.
- Environment — the preview follows the dashboard’s active environment toggle; switching it re-keys the conversation against that environment’s catalog.
Demo data vs your catalog
Before you upload a catalog, the playground runs on a labelled demo storefront so you can try the experience with sample data that never touches real systems. Once your own catalog is live, the agent answers from your real catalog and knowledge — but tool calls are not executed here; they run for real only in your embedded widget.