Agentic Voice Protocol v1.0

The website obeys
your voice.

Celot listens to your visitors and takes action — clicks, navigation, search, form-fills — anywhere on your site. Drop it in with GTM or a single script tag.

"Click services"
Parsed Intent
click_element('Services')
DOM Action
→ <a>Services</a>

What visitors can say

Real intents, native browser actions.

"Click services"
Clicks the Services link in your nav.
"Visit product white shirt"
Navigates to the product detail page.
"Search red sneakers under $80"
Fills your search bar and submits.
"Fill name Ugur, phone 0545…"
Maps spoken fields to form inputs.
"Add this to cart"
Detects context, clicks the right CTA.
"Scroll to pricing"
Smooth-scrolls to the matching section.

The protocol

Voice → Intent → Action.

Listen

Sub-200ms transcription captures the visitor's command and isolates intent from noise.

Parse

Our agent maps speech to your DOM — semantic matching on text, role, and ARIA labels.

Act

Native browser events: click, fill, scroll, navigate. No iframes, no overlays.

Integration

One tag. Any stack.

Deploy with Google Tag Manager, a Cloudflare Worker for edge injection, or paste a script tag straight into your <head>. No framework lock-in.

GTM template
CF Worker
Script tag
// Drop on any page
<script src="https://getcelot.com/agent.js"
data-key="clt_live_…" async> </script>
// That's it. Visitors can now talk to your site.

Give your site ears.

1,000 voice commands a month, free forever. No card required.

Create your account →