Model 3: Bring Your Own Room - Chert FaceTime

Your server creates the LiveKit room and mints a participant token for Chert’s bridge. The bridge publishes the FaceTime caller’s microphone and camera and subscribes to the participant identity you name.

No standing credential exchangeEach call uses a single-use, single-room, short-expiry token. Chert never holds your LiveKit API keys; you never hold Chert’s.

Token requirements

canPublish: true
canSubscribe: true
Scoped to one room
Recommended TTL: call duration plus margin, with a two-hour maximum

Inbound flow

FaceTime rings and Chert sends a signed call.incoming event.
You choose a room and respond within five seconds.
Chert answers, joins, binds to the named participant, and bridges both directions.
Hangup emits call.ended; the token expires and nothing persists.

{
  "event_id": "evt_5a…",
  "type": "call.incoming",
  "call_id": "call_7d1a…",
  "line_id": "line_…",
  "media": "video",
  "from_handle_hash": "sha256:…"
}

{
  "action": "accept",
  "livekit_url": "wss://your-project.livekit.cloud",
  "participant_token": "eyJ…",
  "remote_participant_identity": "acme-anam-agent"
}

No response within five seconds is declined fail-closed. Have the named participant joined or joining before acceptance; if it never appears during the readiness window, Chert ends the call with a named failure.

Outbound

POST /v1/calls
{
  "facetime_handle": "+15551234567",
  "media": "video",
  "livekit_url": "wss://your-project.livekit.cloud",
  "participant_token": "eyJ…",
  "remote_participant_identity": "acme-anam-agent"
}
→ 202 { "call_id": "call_8e2b…" }