VoiceMode is an MCP server that brings human-like voice interactions to Claude Code and other MCP clients. It enables natural voice conversations with AI assistants by bridging microphone input to speech-to-text (STT) and text-to-speech (TTS) back to the assistant, using either local OpenAI-compatible STT/TTS services or cloud providers via the OpenAI API. It supports local voice models (e.g., Whisper.cpp for STT and Kokoro for TTS) and offers a low-latency, real-time experience with automatic transport selection, including optional LiveKit-based rooms. The MCP integration makes it easy to add VoiceMode to Claude Code using the claude mcp commands. It includes silence detection to automatically stop recording when you stop speaking, providing a natural back-and-forth conversation without manual prompts. The project provides installation via Astral UV, NixOS flakes, and multiple installation paths, including from source and package managers. It emphasizes privacy by allowing local processing and provides comprehensive documentation and guides (Getting Started, LiveKit, Whisper, Kokoro).