Qwen Audio Agent

Qwen Audio Agent is a real-time voice runtime designed to keep an AI assistant conversationally present while it performs longer tasks. It supports full-duplex speech, natural interruption, and continuous multi-turn dialogue. Direct questions are answered immediately, while tool-intensive work can be delegated to a background coding agent without ending the voice session. Users can start several independent tasks, check progress, cancel work, and continue discussing returned results. The system can reuse an agent’s existing tools, MCP servers, skills, configuration, and authentication through ACP-based integrations. Interfaces include a browser UI, terminal UI, and a macOS desktop orb with an integrated local gateway. Local profile, memory, and task-state files preserve continuity across sessions while configured audio and backend services handle model processing.