Model 2: Hosted Sessions

Chert hosts the room; your participant joins it.

For teams without LiveKit infrastructure that want their own humans or software on FaceTime calls, including a “FaceTime us” button.

How it works

Chert creates a LiveKit room, bridges FaceTime into it, and returns a short-lived, room-scoped participant token. Your client joins with a standard LiveKit SDK. Chert mints the token; you join Chert’s room.

Outbound session

POST /v1/sessions
{
  "facetime_handle": "[phone]",
  "media": "video",
  "mode": "human"
}

→ 201
{
  "session_id": "sess_9f2c…",
  "call_id": "call_7d1a…",
  "livekit_url": "wss://chert-….livekit.cloud",
  "participant_token": "eyJ…",
  "expires_at": "2026-08-24T18:00:00Z"
}

import { Room } from "livekit-client";

const room = new Room();
await room.connect(livekit_url, participant_token);
await room.localParticipant.enableCameraAndMicrophone();

room.on("trackSubscribed", (track) => attach(track));

Keep credentials server-side

Only the participant token reaches the browser. Your Chert API key stays on your server.

Inbound

Set the line to customer_decides. A call.incoming webhook asks you to respond with { "action": "accept" } or { "action": "decline" }. Acceptance includes the session payload.

Ending

Call DELETE /v1/sessions/{id}, or let either party hang up. Chert emits call.ended and the token expires.