Voice runtime

Voice that stays attached to the work, not outside it.

WorkSwarm treats voice as part of the same operating surface as chat. Calls, transcripts, checkpoints, and post-call actions all land back in the thread so the system stays coherent after the conversation ends.

Why voice is different here

Speak in the composer

Users can switch the chat composer into voice mode and talk to an agent with live transcription and barge-in support.

Receive inbound calls

A tenant-bound phone number can route inbound customer calls into a specific agent and thread context.

Place outbound calls

Agents can place outbound calls with approvals, write the transcript back into the thread, and trigger follow-up work.

Keep one conversation record

Voice does not fork the operating surface. The transcript, tool actions, and post-call summary stay in the same work thread.

The voice story is continuity, not novelty.

The right enterprise claim is not that WorkSwarm can speak. It is that voice becomes another path through the same audited, governed, action-capable system rather than a disconnected calling product.