Developers Attention! Here’s how to add telephony to a Gemini Live Agent with Gemini 3.1 Flash Live, Twilio and Google Cloud

Yesterday, Google shared how developers can add Telephony to a Gemini Live agent to make delightful and multilingual AI conversations possible with just a phone call.

Give your Gemini Live Agent a phone number 📞

Our new guide shows you how to add telephony to your voice agent with Gemini 3.1 Flash Live + Twilio + Google Cloud.

Delightful, multilingual AI conversations are now just a phone call away.

Get the deep dive:… pic.twitter.com/v390HAr307

— Google for Developers (@googledevs) April 27, 2026

In the demo video, presented by Thor Schaeff from Google DeepMind, developers are guided on how they can set up the Gemini Live API with Twilio and deploy it to Google Cloud Run, so anyone can call their Gemini voice agent from a regular phone number.

There are end-to-end examples on GitHub, like the Gemini Live Gen AI Python SDK example that has the Twilio handler built in. The integration works through the WebSocket API.

Developers can build the integration with the help of Coding Agents. If you are using Antigravity, Cursor, or Claude Code, or any other coding agent, you can install the Gemini Live API coding skills, and that will help you in building with the Live API.

In the demo video, the build utilises FastAPI and WebSockets to proxy real-time audio between Twilio and Gemini. The server includes specific endpoints to manage incoming and outgoing calls, using Twilio’s Media Streams to facilitate the connection to the Gemini model.

Also, since Gemini outputs 24kHz 16-bit PCM and Twilio expects 8kHz mu-law audio, the integration requires a conversion step to ensure compatibility. Coming to the deployment to Google Cloud Run, the integration is deployed to Google Cloud Run, using Secret Manager to securely handle API keys for both Gemini and Twilio.

The Gemini Live integration is said to support native audio and camera sharing capability is also supported

If developers do not prefer to manage the low-level audio conversion or WebRTC connectivity manually, Thor has advised several partner integrations such as Lifekit, Pipecat, Fish Jam, Vox Implant and Agora that can handle these requirements.