Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.poly.ai/llms.txt

Use this file to discover all available pages before exploring further.

The Web Calling widget puts your agent on your website. Visitors click the launcher, their browser connects over WebRTC, they’re talking to your agent. You set it up under Configure > Web Calling in Agent Studio: name, domain, environment, branding, copy, embed.The rest of these pages cover the four things you’ll actually do with the widget:

Configure

Brand it, write the welcome line, set the agent voice and disclaimer.

Install on your site

One script tag, paste before </body>. Direct HTML or Tag Manager.

Test your widget

Hosted preview link you can send to anyone, no login required.

Troubleshooting

Microphone, network, cross-tab lock, and other call-quality issues.

​
Before you start

  • The web domain(s) where you’ll embed the widget.
  • Access to your site’s HTML or your tag manager.
  • A published variant in Sandbox or Live for the agent that’ll respond.
  • An HTTPS site. Browsers block microphone access on insecure origins.
PolyAI widgets put your agent on your website in minutes. From Configure > Web Calling in Agent Studio, create a widget, brand it, copy the script tag, and ship it.Phone and Chat widgets share the same editor: same styling controls, same content fields, same embed flow. The widget type just determines whether visitors get a voice call or a text conversation.

​
Pick a widget type

Phone widget (Web Calling)

Click-to-call voice on your website. Visitors talk to your agent directly in the browser over WebRTC. No number to dial, no app to install.

Chat widget (Webchat)

Text chat on your website. Visitors type to the same agent that answers your phone calls, with channel-specific styling and safety filters.
You can run both on the same site. Many teams ship a Phone widget on high-intent pages and a Chat widget elsewhere.

​
What’s the same, what’s different

Everything in the editor (widget name, website URL, variant, environment, agent name, primary color, disclaimer, and policy URLs) works the same way for both types.Chat widgets add a few extra editor fields the call card doesn’t need:
  • Header text and header image (logo).
  • Agent avatar shown next to messages.
  • Consent button (“I consent and start chat”) before the conversation starts.
  • Bubble prompt style defaults (text and background color) for the speech bubble next to the launcher.
Beyond the editor, runtime behavior differs by modality. Phone has microphone permission, mute, call states, and a cross-tab call lock. Chat has transcripts, suggestion chips, attachments, and richer launcher controls (auto-open, render mode, animation, iOS WebView). Both expose a global JavaScript API on the host page so you can drive them from your own code.For runtime and SDK detail, go to the product page:
  • Web Calling: call states, mic permission, cross-tab lock, and the PolyphoneAPI reference.
  • Webchat: chat panel, suggestion chips, and the WebchatAPI reference.
For Chat-specific script-tag attributes (data-platform, data-render-mode, data-auto-open, data-show-icon, data-show-header), see Install on your site.

​
Prerequisites

Before you create a widget you’ll need:
  • The web domain(s) where you’ll embed it.
  • Access to your website’s HTML or your tag manager.
  • A published variant for the agent that will respond.
  • An HTTPS site. Required for Phone (microphone access), recommended for Chat.
For Chat widgets, complete Chat configuration (LLM, greeting, safety filters) before creating the widget.

​
Next steps

Configure widget

Brand it, write the copy, set up consent.

Install on your site

One script tag, live in minutes. Direct HTML or Tag Manager.

Test your widget

Hosted preview link for stakeholder review (Phone widgets).

Troubleshooting

Microphone, network, CSP, and rendering issues.
Last modified on May 18, 2026