Skip to main content
DocsGetting startedQuick start
Getting started

Quick start

Sign up, build your first assistant, and paste the snippet on your site — typical first pass takes about five minutes.

Last updated 2026-05-23

Five steps from a fresh account to a working widget. If you already have an assistant, skip to step 5.

  1. Create your account

    Go to /auth/register, enter your email and a password (or use “Continue with Google”), and check your inbox for the verification link. Full walkthrough: Account.

  2. Create an assistant

    From the Dashboard, click New assistant, give it a name (visitors see this in the chat header) and an internal description, then save. Details: Dashboard.

  3. Add at least one source

    Open the assistant's Sources tab and drop in a PDF, Word doc, Markdown file, or paste a URL to crawl. The build pipeline starts automatically: parsing → chunking → embedding → publishing. You can watch the progress bar on the overview tab. Details: Sources.

  4. Curate scenarios and publish

    Open Scenarios. Type the questions visitors really ask; the app auto-generates each answer from your sources and a few rewording variants. When you're happy, hit Publish. Details: Scenarios.

  5. Paste the snippet

    Open the Embed tab to copy your assistant's snippet (it carries your data-assistant-id). The generic form looks like this:

    <script
      src="https://github.com/pavelsima/zupport-embed/releases/latest/download/embed.js"
      data-assistant-id="YOUR_ASSISTANT_ID"
      defer
    ></script>

    Drop it just before </body> on any page. Platform-specific walkthroughs (Shopify, WordPress, Webflow) are on Embed (per-assistant) and Frameworks.

It takes a minute on first load
The first visitor on a powerful desktop downloads ~570 MB of model weights for the in-browser LLM; everything is cached in IndexedDB afterwards. Phones skip the model entirely and run on scenarios. See Browser support for the full device matrix.

Next