Each brand or site gets a WaveAI collection: its own public key, knowledge base, branding, message quota and allowed origins.
The flow
- A visitor opens the widget and types a question.
- WaveAI searches your knowledge entries for relevant passages.
- The assistant answers using that context, in your chosen tone and reply length.
- If context is thin, or the visitor asks for a person, human-handoff connectors appear.
Grounded means
Grounded means the reply should stay tied to content you uploaded. WaveAI is configured for retrieval-first answers. Gaps should produce an honest "I do not know" plus a route to your team, not confident fiction.
Embed anywhere
The widget is a small script on your pages. It loads asynchronously, respects your accent and launcher position, and only talks to collections that allow your site origin.
Q Can one account run several widgets?
Yes. Use separate collections per brand or site, each with its own key and knowledge.
Q Does the widget need cookies?
The embed is designed for customer sites with normal third-party script loading. If a visitor blocks scripts entirely, the widget cannot run.
Was this helpful?
Thanks. Glad it helped.
Sorry it missed the mark. Visit support or use the WaveAI widget on this site for a quick follow-up.
Prefer a human? Support centre · WaveAI launcher on this page