# Qwen Audio Agent

A realtime voice runtime that keeps an AI agent conversing naturally by voice while it works, so it can keep talking, thinking, and calling tools without breaking the conversation.

Qwen Audio Agent is a realtime voice runtime built so a conversation with an AI agent does not stall while the agent is looking something up, calling a tool, or working on a longer task. It keeps the agent present and responsive by voice throughout, telling the user when a task is complete instead of leaving them waiting after a single reply.

It integrates with existing coding-agent backends over an agent client protocol and agent-to-agent integration, reusing an agent already installed where possible, and adds a desktop conversation panel, remote mobile access, and customizable pet-style avatars linked to the agent's runtime state. It supports fully local voice activity detection, speech-to-text, and text-to-speech for offline operation, alongside example applications such as customer service and a smart cockpit assistant.

It is written in JavaScript, available in Chinese and English, with a published technical report.

Website: https://qwenaudio.github.io/qwen-audio-agent/