Introduction

Why chans?

You focus on your AI. We handle the voice.

Building voice into AI apps is hard – WebRTC, speech recognition, text-to-speech, latency optimization. chans handles all of it so you can ship faster.

• Simple – One SDK, voice AI in 5 minutes. No WebRTC expertise needed.
• Flexible – Use our built-in AI or bring your own. Swap providers via config, not code.
• Your data – Export anytime. Self-host when you're ready. No lock-in.
• Open source – Transparent and auditable. See exactly what runs your voice stack.
• Production-ready – Memory, error handling, MCP tools included.

Start Fast, Graduate to Self-Hosted

Managed Service → Self-Hosted
   (zero setup)    (full control)
        Same SDK, same code

Prototype quickly on our managed service. When you need full control, self-host on your infrastructure – no rewrites, no migration headaches.

How It Works

import { useVoiceAgent } from "@ai-chans/sdk-react"

function App() {
  const { state, connect } = useVoiceAgent({ agentToken: "agt_xxx" })
  return <button onClick={connect}>{state === "idle" ? "Talk" : "Listening..."}</button>
}

That's it. Your users can now talk to your AI.

Two Ways to Build

Enhanced Mode

Get started instantly with built-in AI:

1. Create an agent in the dashboard
2. Set a system prompt
3. Connect with the SDK

No backend required. Conversation memory and MCP tools included.

Passthrough Mode

Keep your existing LLM pipeline:

1. We transcribe speech → send to your webhook
2. You process with your AI
3. Return text → we speak it back

Full control over the AI. We just handle voice transport.

Get Started