An AI voice agent lives on your website.

A talk button sits on your website. Visitor clicks it, grants mic permission, and they're instantly in a real-time voice conversation with your AI agent — powered by WebRTC, running directly in their browser. No downloads, no phone number, no transfers to voicemail.

The whole thing takes about three seconds from click to conversation.

The voice agent isn't pulling generic answers from the internet. It's trained on your business — your services, your pricing, your hours, your policies. Upload your documents and it learns from those. Ask it about your emergency plumbing service at 2am and it answers accurately, specifically, like a team member would.

Choose from over 250 natural-sounding voices powered by Rime AI — filter by gender, language, accent, and model. Preview any voice before you select it. Then fine-tune how the agent behaves.

• Voice speed control — slow it down or speed it up to match your brand tone
• Custom welcome and goodbye messages
• Idle detection with up to 3 retry messages before the agent ends the call
• Interruptible mode — let callers cut in naturally, or let the agent finish
• Max call duration limits to prevent runaway conversations
• Caller memory — toggle on to remember repeat visitors across sessions

The voice agent can do more than talk. It carries the platform's own tools into the conversation — booking, ordering, and lookups happen against your real data, live on the call. No third-party calendar to wire up and no separate booking software: availability comes from your business hours and slot capacity, and an order placed by voice lands in the same list as a website order.

Need it to reach another system? The webhook tool fires custom fields to any endpoint you have, and MCP servers (below) connect it to anything with an API. That's the integration story: native where we can, webhook and MCP where you need it.

• Native booking — checks real availability and books the appointment live on the call, reschedules, and cancels
• Takes orders from your catalog — browses the menu or products, prices the options, places the order with delivery fees included
• Answers with live data — service prices, job status ("where is the technician?"), loyalty points, upcoming appointments
• Email tool — sends a follow-up email to the caller during or after the conversation
• Webhook tool — triggers any external system with custom fields when something happens on a call

For businesses that need custom integrations, the voice agent supports MCP (Model Context Protocol) servers. Use the global MCP servers we provide, or add your own — SSE or stdio transport, with authentication, environment variables, and tool filtering.

This means the voice agent can talk to your CRM, your inventory system, your internal APIs — anything you can expose through an MCP server.

After every call, the agent can generate a summary — with sentiment analysis and custom structured data fields you define. Need to extract "service requested" and "urgency level" from every call? Define those fields and the agent extracts them automatically.

• AI-generated call summaries with custom prompts
• Sentiment analysis (positive / neutral / frustrated)
• Custom structured data fields — define what to extract from each call
• Summary delivery via email — get summaries sent to any address after each call
• Summary delivery via webhook — push structured call data to Zapier, your CRM, or any endpoint

The agent stores memories of previous conversations. A visitor who called last week and asked about your pricing gets a personalized experience this time — not the same generic greeting. It knows their name, their question history, their context.

This isn't a gimmick. It's the difference between a business that feels attentive and one that feels forgettable.

Your admin dashboard shows every voice conversation — full recording, transcript, and an AI-generated summary. You can review any call in seconds without listening to the whole thing.

• Full call recordings
• Complete transcripts
• AI-generated call summaries
• Sentiment scores (positive / neutral / frustrated)
• Duration and timestamp for every call

When a visitor shares their name, email, or phone during a voice conversation, that information flows directly into your Leads Inbox. No manual entry. No missed contacts. Every conversation that captures information creates a lead — tagged as Voice Agent, timestamped, and ready for follow-up.

Voice AI infrastructure is expensive to run. We absorb all of it. Per-call costs, WebRTC infrastructure, TTS and speech recognition — all bundled into your monthly plan. You use it as much as you want. The bill never changes.