Web channel

Website Assistant

Add a conversational AI assistant to a business website with one script tag. Conversations land in the project Inbox and reuse the tone, context, and rules already configured.

The Website Assistant uses a browser-safe wgt_pk_... key. It is separate from server keys gliiz_... and cannot publish, generate visuals, or read account data.
What it unlocks
ConversionAI chatVisitors can ask natural-language questions without leaving the website.
InboxWeb channelEach exchange is stored in the Inbox with platform = web.
BrandProject contextThe bot reuses the project's tone, information, and FAQ.
Install1 scriptNo heavy integration: create a widget key, allow domains, then copy a snippet.
Where to generate and manage the key

The embed component lives in the Public API (v1) section, in the same place as other project keys. There is no second hidden page to remember.

1ProjectOpen the target project.
2Settings / IntegrationsOpen the Public API (v1) section.
3Create keyChoose Website Assistant, then add the allowed domains.
4Active keysFind the snippet later on the Website Assistant key row.
5SettingsEdit domains, greeting, and monthly cap without regenerating the key.
Code to copy

Replace wgt_pk_...with the key shown in Gliiz. The website domain must be listed in the key's allowed domains.

html
<script
  src="https://gliiz.com/widget.js"
  data-gliiz-site="wgt_pk_..."
  data-gliiz-lang="en"
  async
></script>
The script detects the Gliiz origin from src, initializes the widget, then calls /api/widget/bootstrap and /api/widget/message over CORS. Change data-gliiz-lang to en, fr, es, pt, or ar for the website language.
Security and limits
Keywgt_pk_...Browser-visible, but scoped to widget chat only.
Domainsallowed_originsEvery call checks the website origin before replying.
PlansStarter+Free is excluded. Starter, Pro, Agency, and Enterprise can use it.
Per-key capwidget_monthly_limitThe optional spend guard is counted on the specific widget key.
UsageReplies/monthThe key list shows generated replies this month per key.
MediaPreparedImage and voice are prepared in the UI; v1 replies remain text-only.
Documentation Feedback

Was this documentation helpful?