qwenaudio/qwen-audio-agent

Qwen Audio Agent is a realtime voice runtime that lets you talk to an AI coding assistant out loud instead of typing, while the assistant keeps working on tasks in the background and stays present in the conversation the whole time. Rather than making you wait in silence while it looks something up, calls a tool, or finishes a task, it keeps talking with you, and when a task is done it tells you out loud that it is ready. The tool sits on top of existing coding agents rather than replacing them. You pick a backend agent you already use, such as OpenCode, Qwen Code, Kimi Code, Codex, or Claude Code, and Qwen Audio Agent handles the voice layer around it. You can hold a full duplex conversation, meaning you and the assistant can both speak and be interrupted naturally, much like a real phone call, rather than taking strict turns. Conversation and background work happen at the same time. You can ask the assistant to start a task, keep chatting about something else while it works, check on progress, or cancel it, and when the task finishes the result comes back into the same ongoing conversation so you can ask follow up questions or request changes. You can also start multiple independent tasks that run at once and are tracked separately. It is available as a web interface, a terminal based text interface, and a desktop floating orb for macOS, Windows, and Linux. The desktop app can remember details about you across sessions, building up personalization over time without you needing to repeat yourself. To install it you need Node.js version 22.22.2 or newer, or 24.15.0 or newer, along with npm 10 or later, and it can then be installed globally with a single npm command. Building from source and installing from GitHub are also documented, along with the setup steps for obtaining an API key needed to use its voice features.