omni:session, an audio transport, and a playback path that handles
interruptions. A WebSocket connection alone does not capture or play audio.
Same runtime for UI and API agents
Create a saved agent in the console or withPOST /v1/agents, then pass its returned agent_id
as session_label when opening /v1/omni. The saved voice, instructions,
knowledge and tools are loaded for the call. You can also configure sessions
inline without saving a profile. The endpoint and omni:session scope stay the same.
One endpoint instead of a voice cascade
Omni manages speech recognition, reasoning, speech generation, and turn-taking within the session. Your application still owns capture, playback, and any client or phone actions. Use the product directory if you only need STT or TTS, or want a framework to own the conversation loop.Connect
This Node.js wire example opens a session without a saved profile. Keep the API key on your server. For an SDK-managed connection, use the official SDK guide; browser clients need a short-lived token from a server broker or the hosted widget.configured control frame before sending non-silent caller audio. tools is
the accepted count. After that, send PCM16 as 0x01 || audio_bytes. Follow the browser tutorial for capture and
playback and the wire protocol for framing. Preserve
structured transcript roles. If updating an older
client, use the migration guide.
Choose how to build
Settings and customization
Prompts, voices, languages, conversation behavior, and per-call overrides.
Launch an Agent
Create, test, and put a Call Now button on your site without a token broker.
Create agents via API
Store a profile, bind knowledge and tools, open Omni with session_label.
Build the browser client
Capture the microphone, mint a short-lived token, handle frames, and play audio.
Put Omni on a phone call
Bridge Twilio media, barge-in, DTMF, and call-control events.
Read the wire protocol
Binary audio framing, lifecycle events, tools, and session controls.