Terri answers and places phone calls through a live speech-to-speech session.
With voice_inbound enabled, calls to your number are answered by the agent. It greets the caller and responds to what they say, guided by your system prompt.
Agent replies stream through a live speech-to-speech connection, so callers hear responses as soon as the model starts speaking.
Terri handles both DTMF keypad input and spoken answers, including phone menus and verification prompts.
When the far party needs the account owner, Terri conferences a number from handoffPhoneNumbers into the same live call. Terri stays on the line so all three can hear each other. Blind transfer (voice_transfer) still exists for dump-and-dial handoff.
With voice_outbound enabled, the agent can place calls — but only to destinations your outbound policy allows and only while billing is in good standing.
Outbound call creation and classic voice actions pass maxCallDurationSeconds to Twilio. The normal Realtime inbound stream path does not currently enforce that field when the stream starts.