Qwen Audio Agent
A realtime voice layer on top of Claude Code, Codex and OpenCode that keeps talking and listening while background tasks run in parallel
Medium risk
We rate an entry medium when the tool runs code, makes network calls or reads project files. Check what exactly it does before installing.
Why this level
- Requires a cloud DashScope key and continuously listens to voice input through a separate gateway
Install
In your terminal, with SkillFoxx CLI
npx skillfoxx add cli/qwen-audio-agentDetects the agents on your machine, checks the risk and pins the version.
Other ways to install
Install the tool
npm install -g qwen-audio-agentInstall: npm install -g qwen-audio-agent. Set up the key with qwenaudio config (needs DASHSCOPE_API_KEY), then start the gateway with qwenaudio and, in another terminal, the interface with qwenaudio tui.
Other ways from the author
npm install -g qwen-audio-agentRequires Node.js 22.22.2+ or 24.15.0+.
This is third-party code. Review the repository files before installing.
What it does
Qwen Audio Agent runs a voice gateway that plugs into an already installed coding agent as a backend over the ACP protocol. The conversation keeps flowing while the agent searches the web, calls a tool or runs a background task: the voice keeps responding, and the background task's result returns to the same conversation once it's ready. It ships a terminal UI, a web UI, and a desktop app with a floating voice orb for macOS, Windows and Linux, plus long-term memory across sessions.
Who it is for. For developers who want to talk to their coding agent by voice instead of typing, without losing access to its tools and MCP servers.
Good fit when
- You want to drive Claude Code, Codex or OpenCode by voice while your hands stay free for other work
- You need a voice assistant that doesn't go silent while a long background task runs
- A floating desktop voice orb suits you better than typing into a terminal
Not a fit when
- You have no Alibaba Cloud Bailian account for a DashScope key and won't get one
- You don't have the required Node.js version, or running a separate voice gateway isn't convenient
Example request
Start the voice gateway and connect it to my current Claude Code sessionLimitations
Requires Node.js 22.22.2+ or 24.15.0+ and a DashScope API key from Alibaba Cloud Bailian for the default voice frontend. The voice frontend and the backend agent are configured separately: Qwen Audio Agent doesn't write code itself, it hands tasks to the installed agent over ACP.
How to disable. Uninstall the package: npm uninstall -g qwen-audio-agent, and stop the gateway and TUI processes if they were running in the background.
Security check
- Requires a cloud DashScope key and continuously listens to voice input through a separate gateway
README in short
The README describes Qwen Audio Agent as a realtime runtime that keeps the agent voice-present during background work. Version 2.0 rebuilt orchestration around a unified client protocol with ACP and A2A backend integration, added new voice and video model support, and a desktop conversation panel with remote mobile access. Example scenarios include a smart cockpit, a digital human and customer-service setups. Installed via npm, configured with qwenaudio config, and run as a separate gateway and interface command.
FAQ
Which agents does this work with?
Agents that speak the ACP or A2A protocol; the README and the repository topics name Claude Code, Codex and OpenCode.
Is DashScope mandatory?
The default voice frontend uses DashScope, but the README also mentions other cloud and self-hosted voice frontend options.
Related
A self-hosted knowledge base with block-level references and a built-in MCP server for connecting AI agents to your notes
A CLI for every Google Workspace API with JSON output and agent skills: Drive, Gmail, Calendar, Sheets and more
Local search over Markdown notes, docs and meeting transcripts: keywords, semantic search and reranking, with an MCP server
A task manager for AI-driven development: breaks a PRD into dependent tasks and guides the agent through them via MCP or CLI