Get started with live agents¶

The quickstarts run your agent in adk web, which ships a browser client that already captures the microphone, plays the agent's replies, and renders the transcript. You write the agent and pick a model; there is no client code in the way.

Your agent needs a model that can hold a two-way streaming connection. See Supported models for the current list and how to configure one.

Choose your language¶

Python

Set up ADK, build a voice agent, and talk to it in adk web.

Python quickstart
Java

Set up Maven, build a voice agent, and run it in adk web or a custom audio app.

Java quickstart

Next steps¶

Configuration set the voice, language, transcription, and turn detection.
Tools give the agent tools it can call mid-conversation, including ones that stream results back while they run.
Sessions and Events the run_live() loop and everything it hands back.
Evaluation score voice conversations before you ship.
Build a custom server adk web is a development client, so this is how you run a live agent behind your own server and client.