POST /api/v1/calls with an explicit line and your call-scoped BYOR
credentials, just like Outbound.
These existing capabilities are experimental and enabled separately per line.
Ask Chert to confirm access before integrating. A line enabled for ordinary
outbound calls does not automatically support groups or links. Supervised group
media acceptance is still pending; availability is not a delivery guarantee.
Start a group call
In the outbound request, replacedestination and destination_type with
recipients. Supply 1–31 distinct remote handles; use at least two for a group:
direction, mode, line_id, media, and all
LiveKit fields from the outbound example.
Use phone_number inside the recipients array, rather than the single-recipient
destination_type: "phone" spelling. Do not combine the array with a destination
or a FaceTime link. Reordering the same recipient set does not create a new
operation when you reuse its idempotency key.
The line invites everyone into one conversation. Invitations are not proof that
anyone answered. Unanswered invitations and partial joins are possible.
Join a FaceTime link
Use these destination fields in the outbound request:Media, capacity, and ending
- One group reserves one line/call slot, regardless of participant count.
- Group audio is mixed. There are no per-person tracks, speaker attribution, participant-count, or moderation APIs in this contract.
- Video requires the line’s separate outbound video capability. Caller capture does not guarantee coverage of every group tile.
- Wait for
bridge_readybefore the agent speaks. Use the normal end-call API and observe terminal cleanup before reusing the line. - Reuse the original idempotency key after an uncertain create response; do not submit a second group or join with a fresh key to work around uncertainty.

