Natural voice conversations for AI assistants. Voice Mode enables real-time, human-like voice interactions with Claude, ChatGPT and any other MCP-capable LLM via the Model Context Protocol (MCP).
https://github.com/mbailey/voicemodeStop typing. Start talking. Voice Mode transforms your Claude, ChatGPT, and other LLM interactions from text-heavy sessions into natural voice conversations through the Model Context Protocol.
You're spending hours typing complex prompts, copying responses, and managing context in text-based AI interactions. Meanwhile, you could be having natural conversations that flow like talking to a colleague - asking questions, getting immediate spoken responses, and iterating on ideas in real-time.
Voice Mode eliminates the friction between your thoughts and AI assistance. Instead of "Let me type this complex question about architecture patterns," you simply say "Explain the trade-offs between microservices and monoliths for my e-commerce project" and get an immediate spoken response.
Real-time iteration speed: Voice conversations happen at the speed of thought. No more typing, waiting, reading, then typing again. Ask follow-up questions naturally and get immediate context-aware responses.
Hands-free coding sessions: Review code, discuss architecture, or debug issues while keeping your hands free for other tasks. Perfect for pair programming sessions or when you need to reference multiple screens.
Natural context building: Voice conversations naturally build context better than text exchanges. The back-and-forth flow helps both you and the AI maintain conversation thread and dive deeper into topics.
Accessibility and comfort: Some developers simply think and communicate better verbally. Voice Mode accommodates different working styles and accessibility needs.
Voice Mode integrates seamlessly with your existing MCP-enabled tools:
# Claude Code - one command setup
claude mcp add --scope user voice-mode uvx voice-mode
# Any MCP client - standard configuration
{
"mcpServers": {
"voice-mode": {
"command": "uvx",
"args": ["voice-mode"],
"env": {
"OPENAI_API_KEY": "your-openai-key"
}
}
}
}
Works immediately with Claude Desktop, VS Code, Cursor, Zed, Continue, Windsurf, and any other MCP-compatible environment. No custom clients or complex setups required.
Local processing options: Run Whisper.cpp for speech-to-text and Kokoro for text-to-speech entirely on your machine. Zero cloud dependencies for sensitive projects.
OpenAI-compatible APIs: Seamlessly switch between providers or implement custom routing. Your voice data flows through the services you choose and trust.
Transparent routing: Deploy your own API gateway to route requests based on cost, latency, or privacy requirements without changing Voice Mode configuration.
Code review sessions: "Walk me through this React component's performance implications" while looking at your IDE. Get spoken explanations while keeping your hands free to navigate code.
Architecture discussions: "I'm designing a real-time chat system. What are the WebSocket vs Server-Sent Events trade-offs?" Natural back-and-forth that builds on previous context.
Debugging conversations: "This async function is causing race conditions. Let's talk through the execution flow" - perfect for rubber duck debugging with an AI that can actually respond.
Learning sessions: "Explain dependency injection patterns, but pause after each example so I can ask questions." Interactive learning that adapts to your pace.
Multiple transport modes:
Audio format optimization: PCM for real-time streaming performance, with automatic fallback to supported formats. Configurable quality settings for different use cases.
Cross-platform reliability: Tested on Ubuntu, Fedora, macOS, and Windows/WSL with clear setup instructions for each platform's audio stack.
Developer-friendly debugging: Built-in audio diagnostics, debug mode with file saving, and comprehensive troubleshooting guides.
The fastest path to voice-enabled AI conversations:
# Install dependencies (varies by platform)
curl -LsSf https://astral.sh/uv/install.sh | sh
export OPENAI_API_KEY=your-openai-key
# Add to Claude Code
claude mcp add --scope user voice-mode uvx voice-mode
# Start voice conversation
claude converse
# Then say: "Let's have a voice conversation"
For local privacy, add Whisper.cpp and Kokoro TTS:
# Point to local services
export STT_BASE_URL="http://localhost:2022/v1" # Local Whisper
export TTS_BASE_URL="http://localhost:8880/v1" # Local TTS
Voice Mode automatically detects and uses these endpoints while maintaining the same OpenAI-compatible interface.
Voice interfaces are becoming standard in professional software development. Voice Mode positions you ahead of this trend by integrating voice capabilities directly into your existing AI workflow rather than requiring separate tools or platforms.
The MCP integration means you're not locked into a specific AI provider or development environment. Voice Mode works wherever MCP works, making it a future-proof addition to your development toolkit.
Your AI conversations become more natural, efficient, and productive. Instead of fighting with text interfaces, you're having actual conversations that feel like working with a knowledgeable colleague who happens to be an AI.
Get Voice Mode: uvx voice-mode | getvoicemode.com | GitHub