For builders
Build your own AvA
Plugins extend the assistant. The Realtime API goes a level deeper — it is the same agentic backend AvA itself runs on. Bring your own interface and build something entirely different on top.
Realtime voice in and out
Sub-second, full duplex, with barge-in handled for you.
Vision alongside audio
Screen and camera context in the same session as the conversation.
Live transcripts
Streaming text for both sides of the conversation as it happens.
Tool calls to your backend
The model asks; your code answers. You keep the business logic.
WebSocket and WebRTC
Whichever transport suits your client. Both are first class.
Your UI, your product
A companion, a tutor, a kiosk, an in-car assistant. The engine does not mind.
Realtime API
The streaming voice and vision protocol. Message types, session lifecycle, tool-call contract.
Read the protocol docsREST API & OpenAPI
Assistants, plugins and knowledge base over plain HTTP. Spec, scoped keys, and a live playground.
Open the API reference