Chat Widget

The chat widget puts your Tridan AI chat agent in the corner of every page on your website. It is embedded as an iframe plus a small loader script, answers from the content you train it on, and captures leads while it works — no code changes to your site beyond one snippet.

How it works

From an empty agent to a live widget in four steps.

  1. 1

    Create an agent

    In the Tridan dashboard, create an AI agent and upload the content it should know: documents, FAQs, and website pages.

  2. 2

    Configure the widget

    Set the appearance, intake form, nudges, and the origin allowlist in the agent's Embed tab, then copy your API key.

  3. 3

    Embed it

    Paste the iframe snippet plus the widget-context.js script before the closing </body> tag on your website.

  4. 4

    It starts answering

    The widget goes live immediately. Review conversations, leads, and feedback in the dashboard.

For the exact embed code (plain HTML, React, and WordPress), see the widget setup guide.

Features

Everything below is configured per agent in the dashboard — no redeploys on your website needed.

Page-aware answers

The optional widget-context.js loader lets the agent read the host page, so it can answer questions about whatever the visitor is currently viewing.

Pre-chat intake forms

Collect a name, email, or any custom fields before the conversation starts, so every chat begins with the context you need.

In-chat lead collection

The agent can ask for contact details naturally during the conversation and store them as leads.

Page nudges

Show proactive messages on specific pages to invite visitors into a conversation at the right moment.

Feedback collection

Visitors can rate answers, giving you a direct signal on where the agent's knowledge needs improvement.

Per-user threads

Each visitor gets their own conversation thread, so returning visitors can pick up where they left off.

Origin allowlist

The widget only loads on domains you approve, preventing your API key from being embedded on other sites.

Content moderation

Built-in moderation filters abusive or off-policy input before it reaches your agent.

Themeable appearance

Adjust colors and appearance so the widget matches your website's branding.

Authenticated mode

Optionally pass a user token so the agent knows which logged-in user it is talking to.

Multi-language

The agent replies in the language the visitor writes in — no extra configuration required.

4 leading AI models

Choose the model behind your widget from OpenAI, Anthropic, Meta, and DeepSeek.

Next steps

Launch your chat widget

Create an agent yourself, or book a call and we will configure the widget with you.

Last updated · Reviewed by Nemanja Milivojevic, Founder