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

# Introduction

> FaceTime, made programmable.

Chert turns FaceTime into programmable infrastructure. Run AI video personas on real FaceTime lines, embed live FaceTime calling in your product, or bridge FaceTime into LiveKit infrastructure you already own.

<CardGroup cols={3}>
  <Card title="Assistants">
    Model 1

    Chert runs the avatar, voice, models, and real-time conversation for you.
  </Card>

  <Card title="Hosted Sessions">
    Model 2

    Chert hosts the room; your human or software participant joins it.
  </Card>

  <Card title="Bring Your Own Room">
    Model 3

    Chert joins your LiveKit room as a dedicated FaceTime gateway.
  </Card>
</CardGroup>

<Note>
  **Product availability**

  **Assistants are generally available.** Sessions, outbound calling, webhooks, and transcript or summary retrieval are in design-partner rollout. Draft contracts may receive minor changes before GA.
</Note>

<h2 id="how-it-works">
  How it works
</h2>

Every Chert FaceTime line is backed by managed infrastructure that answers and places real FaceTime calls and converts them into standard real-time media. The caller's audio and video become tracks in a LiveKit room; audio and video published into that room are delivered back to the caller.

The three models differ only in who is on the other side: Chert's AI persona, your participant in a Chert-hosted room, or Chert's bridge in a room you host.

<h2 id="fail-closed">
  Designed to fail closed
</h2>

Chert never answers a call it cannot serve properly. The system verifies its media path, avatar, and agent readiness before and during operation. If a check fails, inbound calls are declined instead of being answered into silence or broken video.

<Tip>
  **The operational promise**

  The worst experience a caller should have is a busy signal.
</Tip>