How is the assistant integrated into my website?

Integration is done via a short code snippet that fits common websites, CMS and shops. It consists of exactly one script line with the public key of your assistant; colour, side, vertical offset and language can be set directly on the snippet and override the setting on the server.

How this works technically and which options exist for the display is described on the integration and embedding page. As a rule, a single snippet is enough to make the assistant available on all pages.

The loader opens a shielded area on your page, draws the trigger inside it and only attaches the frame after the page has fully loaded; the loader itself never calls the network. A second snippet on the same page returns silently, so a window embedded twice does not appear twice. Without JavaScript nothing appears, and there is no fallback path for that.

The frame carries its own opaque origin: the window does not read along on the host page, and the host page cannot read the content of the window. Commands and the conversation ID pass between the two; anyone who knows the ID can retrieve the history, including scripts on the host page. Delivery happens on two paths, via our widget domain or via a path on your own domain that you pass on to us. Updates reach your page without action on your part because the files carry version names derived from their content; an unknown name is an error and not a silent fallback. The steps are in the documentation.