> ## Documentation Index
> Fetch the complete documentation index at: https://docs.mrassistant.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Web and voice

> How customer surfaces and realtime voice fit together

MrAssistant AI is built around **realtime, conversational** experiences. In most B2C setups, the customer interacts through your **app or website**, while your services coordinate **sessions**, **agent configuration**, and **downstream integrations**.

## Mental model

1. **Configuration** — You define how the assistant should sound, which languages apply, and which high-level capabilities are in scope.
2. **Sessions** — A session is the unit of a customer attempt: one guided flow or conversation.
3. **Media path** — Audio and timing depend on the **realtime stack** you connect; treat that as a product decision alongside UX and support.

## Web and mobile clients

Keep **business rules and integration logic** on systems you operate, not embedded in static client bundles. Clients should focus on **capture, playback, and UX**, while your backend decides what the assistant may do next.

## Regional and reliability planning

Voice is sensitive to **latency** and **distance**. When you move from pilot to production, plan for **where traffic runs** and how you observe quality (drops, silence, user retries).

## Next steps

<CardGroup cols={2}>
  <Card title="Voice and providers" icon="waveform" href="/voice/providers" />

  <Card title="Multilingual" icon="language" href="/voice/multilingual" />
</CardGroup>