Skip to main content

Put your agent on any website

Any Digitorn agent can run as a chat widget on a site you already have — your marketing page, your app, your docs — with a single <script> tag. The chat runs in an isolated iframe (it can't clash with your CSS), and you control how it looks, how it opens, and what it knows about the visitor.

html
<script src="https://digitorn.ai/embed.js" data-key="dk_YOUR_KEY"></script>

That one line gives you a launcher that opens your agent's chat as a full-height panel on the side of the page. Everything else on these pages is optional.

Get your key​

The key (dk_…) identifies one embed of one agent, and usage is billed to you (its owner). Create it in the app:

  1. Open your agent in Digitorn.
  2. Go to the Embed tab.
  3. Create a key and copy the snippet.

You can restrict a key to specific domains, restyle it, and revoke it at any time from the same tab — without ever touching the site it's pasted on.

Two ways to configure​

Every option can be set two ways, and they mean the same thing:

  • On the tag — a data-* attribute, e.g. data-layout="corner". Good when you want the site's HTML to own the choice.
  • On the key — saved in the Embed tab. Good when you want to restyle the widget later without asking anyone to edit the page.

A data-* attribute on the tag always wins over the key, so a value you spelled out in your HTML is never silently overridden.

What's on these pages​

  • Appearance — theme, full colour override, the AI stylist, layout, launcher, sizes.
  • Make the page adapt — let a docked chat reflow your page instead of covering it, the way docs.digitorn.ai does.
  • Behaviour & triggers — open on a delay, on scroll, on exit intent, or from your own button.
  • JavaScript API — control the widget from your code, and tell the agent who the visitor is and what page they're on.
  • Visitors & verified users — each visitor is their own user of your agent, with their own accounts; see them all, and verify the users of your own site.
  • Security & privacy — domain allowlist, and what the agent should and shouldn't trust.