One Widget, Any Role Your Business Needs
A Chat AI Agent is a widget on your website that answers visitors from your own knowledge, takes actions such as booking meetings or creating tickets through integrations, and hands conversations to your team when needed. Tridan AI's widget works as a sales assistant, lead generator or support agent, and visitors can type or talk.
Chat Agent: Website Widget
Type or talk, same conversation
Configure the Agent's Role
- Sales Assistant
- Lead Generator
- Knowledge / Support Agent
- Custom Role
What is a Chat AI Agent for a website?
A Chat AI Agent for a website is a chat widget embedded on your pages that answers visitors from your own documents and knowledge, performs actions such as bookings and tickets through integrations, and hands conversations to people when needed. Tridan AI's agent runs on models from OpenAI, Anthropic, Meta and DeepSeek and knows which page the visitor is on.
Read the chat widget documentationWhy put a Chat AI Agent on your website?
A Chat AI Agent answers visitors instantly, at any hour, from your own content instead of generic web knowledge. One widget can qualify leads, recommend products, resolve support questions and book meetings, and you change its role by rewriting instructions. Visitors can also speak instead of typing, which keeps mobile conversations going.
Any Role, No Code
Switch the chat agent from a sales assistant to a support agent by rewriting its instructions. No developer required.
Type or Talk
Visitors can tap the mic and speak instead of typing. It is faster, more natural and built into the same widget.
Trained on Your Knowledge
The chat agent answers from your docs, FAQs and files, not generic web knowledge, so answers match your business.
Deploy in Minutes
Paste a short embed snippet on your site and the widget goes live with text and voice mode included.
- Visitorasks a question on your site
- Chat agentreads the page and your docs
- Instant answerfrom your knowledge, any hour
- Actionbooks a meeting or opens a ticket
- Your teamhandoff inbox with transcript
How do you add a Chat AI Agent to a website?
You add a Chat AI Agent in three steps: pick a role or write your own instructions, connect your knowledge and the integrations the agent may use, then paste the embed snippet before the closing body tag of your site. The widget loads only on domains you approve, and visitors can start typing or talking immediately.
- 01
Pick a Role
Start from a Sales, Lead Generator or Support preset, or write custom instructions for anything else.
- 02
Connect Your Knowledge
Upload docs and files, and connect the tools the chat agent should use, such as Google Calendar, Zendesk or Jira.
- 03
Embed and Go Live
Paste the embed snippet on your site. Visitors can type or talk to the chat agent immediately.
Configure the Chat Agent for the Job at Hand
Start from a preset built for a common job, or write your own instructions from scratch.
Sales Assistant
Qualifies visitors, recommends plans or products, and pushes toward checkout or a booked demo.
Lead Generator
Captures contact info and intent from anonymous visitors and hands qualified leads straight to your team.
See Lead Generation in depthKnowledge / Support Agent
Answers product and account questions from your knowledge base and deflects routine support tickets.
Custom / Build Your Own
Don't fit a preset? Define any role with plain-language instructions. No code required.
Need it fully custom?Can visitors talk to the chat agent instead of typing?
Yes. Visitors tap the microphone in the widget and speak, and the chat agent replies in a natural voice from ElevenLabs, Deepgram or OpenAI, the same voices used by the Call Center Voice Agent. Switching between typing and talking keeps the conversation context, and no extra install is needed.
Tap to Talk
One tap on the mic switches from typing to speaking, mid-conversation, without losing context.
Natural Voice Replies
The chat agent answers back in a natural-sounding voice, powered by the same voices used on phone calls.
Faster Than Typing
Some visitors prefer to speak rather than type, especially on mobile.
How does a Chat AI Agent compare with a contact form or live chat?
A Chat AI Agent replies within seconds at any hour, answers from your documentation and can book meetings or open tickets through integrations. A static contact form only collects a message for later, and live chat alone depends on an agent being online. Handoff to your team stays available in every case.
Swipe sideways to see the full table
| Criteria | Chat AI Agent | Static contact form | Live chat only |
|---|---|---|---|
| Response time | Seconds, every time | Hours or days, by email | Minutes, when an agent is online |
| Answers from your docs | Yes, from uploaded files and knowledge | No | Depends on the agent's knowledge |
| Actions (bookings, tickets) | Yes, through Google Calendar, Zendesk, Jira and more | No, manual follow-up | Manual, by the agent |
| After-hours | Fully available | Collects messages only | Offline or a message form |
| Handoff | To the handoff inbox with the transcript | Not applicable | Between agents only |
| Best for | Instant answers, lead capture and self-service | Low-volume inquiries | Complex cases with a staffed team |
Choose a Chat AI Agent when visitors expect an answer now and you want the team involved only when it matters.
Where Teams Put the Chat Agent to Work
One widget, many jobs: across your website, app and internal tools.
Support Deflection
Resolve common questions instantly and free your team for harder tickets.
E-Commerce Sales Assistant
Recommend products, answer sizing or shipping questions, and reduce cart abandonment.
SaaS Onboarding Helper
Walk new users through setup and answer product questions inside the app.
Lead Capture on Landing Pages
Turn anonymous traffic into qualified leads before they bounce.
Internal Knowledge Assistant
Give employees instant answers from internal docs and wikis.
Multilingual Websites
Serve visitors in their own language with the same agent, knowledge, and rules.
Where does your chat data live?
Your conversations and knowledge stay inside your Tridan AI workspace, and with the self-hosted option the whole platform runs in your VPC or on-premises with local models through Ollama. Built-in PII detection and anonymization protect personal data in transcripts, and the widget loads only on domains you approve.
Self-Hosted Option
Run the full platform in your VPC or on-premises, including local models via Ollama.
PII Detection
Built-in detection and anonymization of personal data in conversations.
Multilingual by Default
One widget for every language you configure, with no separate setup per locale.
Common Questions About the AI Chat Widget
Can I change the chat agent's role after launch?
Yes. Update the instructions at any time and the chat agent's behavior changes immediately, with no redeployment and no change to the embed snippet on your site. Many teams start with the support preset, then add lead qualification or product recommendations once they see which questions visitors ask. The conversation history and uploaded knowledge stay in place.
Does voice mode work on mobile?
Yes. Voice mode works in desktop and mobile browsers wherever microphone access is available. The visitor taps the mic icon in the widget, speaks, and the chat agent replies in a natural voice. Switching between typing and talking keeps the same conversation, so a visitor can start on the phone screen and continue by text later.
How is the chat agent trained on our content?
You upload documents, FAQs and files to the agent's knowledge, and the chat agent retrieves the relevant passages when it answers. No fine-tuning or machine-learning expertise is needed, and you can add or replace files whenever your content changes. Connected sources such as Notion or Confluence pages can also serve as knowledge.
Do we need developers to embed the widget?
No. Installing the widget means pasting a short embed snippet, an iframe plus an optional widget-context.js script, before the closing body tag of your site. Most teams add it through their CMS or tag manager without engineering time. The widget only loads on the domains you approve in the agent's Embed settings.
Can the chat agent read the page the visitor is on?
Yes. The optional widget-context.js loader lets the chat agent read the host page, so it can answer questions about the product, article or pricing table the visitor is currently viewing. You can also show proactive messages on specific pages to start a conversation at the right moment. Both options are configured in the agent's Embed tab.
Is my data used to train models?
Your uploaded knowledge and conversations stay in your Tridan AI workspace and are used only to answer your visitors. With the self-hosted option, the whole platform, including local models through Ollama, runs inside your own infrastructure, so data never leaves it. Built-in PII detection and anonymization can mask personal data in transcripts before anyone reviews them.
How do I install the widget?
Open the agent's Embed tab, set the appearance, intake form and allowed domains, then copy the embed snippet. Paste the iframe snippet and the widget-context.js script before the closing body tag on your website, or add them through your CMS or tag manager. The documentation walks through each step with examples.
Which actions can the chat agent take for visitors?
Through integrations, the chat agent can book meetings in Google Calendar or Outlook, create tickets in Zendesk or Jira, send email through Gmail, Outlook or SMTP, read knowledge from Notion and Confluence, and continue conversations on WhatsApp. Anything outside its instructions is routed to your team through the handoff inbox with the transcript attached.
Last updated · Reviewed by Nemanja Milivojevic, Founder
See the Chat Agent Handle Your Role
Talk to us about your use case and we'll show you the chat agent configured as a sales, lead generation or support agent, with text and voice included.