Lets you talk with your coding agent out loud: it speaks its replies and listens for yours, through OpenAI or local Whisper and Kokoro.
by Voicemode· MediaUnclaimed · Yours? Claim it
4 images
About
VoiceMode adds natural voice conversations to Claude Code and other agents that run MCP servers. Your agent speaks to you through your speakers, listens through your microphone, stops recording when you stop talking, and treats what you said as your next message. It's an open-source project by Mike Bailey (MIT licensed), made for the moments when your hands or eyes are busy: walking to a meeting, cooking while you debug, or resting your eyes after hours at the screen. What you can do with it - Hold a spoken back-and-forth with your agent while it works - Have it narrate what it's doing without waiting for a reply, or ask several questions in one go - Choose the voice, speaking speed and speech provider - Start, stop, check and read the logs of local Whisper (speech-to-text) and Kokoro (text-to-speech) services - Keep the floor while it does background work, so another agent on the same voice channel doesn't cut in How it connects It runs on your computer as a local MCP server from the voice-mode package on PyPI (uvx voice-mode), with Python 3.10 or later on Linux, macOS, Windows (native or WSL) or NixOS. You need a microphone and speakers. Speech goes to OpenAI (set OPENAI_API_KEY) or any OpenAI-compatible service, or to local Whisper and Kokoro services, which keep it private and work offline. Its installer (uvx voice-mode-install) sets up system dependencies such as FFmpeg and PortAudio and offers to install the local services. Claude Code users can also add it as a plugin from Mike Bailey's own marketplace (mbailey/voicemode). Good to know - Without an OpenAI API key, it needs the local Whisper and Kokoro services running. Their models download the first time they start (about 500 MB for the recommended pair). - It uses local services when they're available; they share OpenAI's API, so it can switch between them. - It's an independent project, not made by Anthropic or OpenAI.
Once it’s added
“Let's talk this through by voice.”
“Narrate what you're doing as you refactor, without waiting for me.”
“Ask me your three setup questions out loud, one after another.”
“Pause our conversation while you run the tests, then tell me how they went.”
Before you add it
Reads only
What you say into your microphone, turned into text · Status and logs of the local Whisper and Kokoro services
Makes changes · your agent can ask first
Speaks its replies through your speakers · Starts, stops, enables or disables local voice services
From people who use it
Step by step
Nothing new reaches your agent until a person has looked at it.
Plugins people install alongside VoiceMode.
No reviews yet. Used VoiceMode? Say how it went.
Write a review“Check whether Whisper and Kokoro are running.”
“Start the local Kokoro voice service and show me its logs.”
Scanned on every change, and again when our rules improve.
CursorSupported
Claude CodeSupportedTell us and a person on the trust team will look at it. Or email [email protected].
Runs on your machine.
claude_desktop_config.json{
"mcpServers": {
"voicemode": {
"command": "uvx",
"args": [
"voice-mode==9.0.1"
]
}
}
}Runs on your machine.
Add to VS Codecode --add-mcp '{"name":"voicemode","command":"uvx","args":["voice-mode==9.0.1"]}'copilot mcp add voicemode -- uvx voice-mode==9.0.1
Runs on your machine.
Add to Cursor.cursor/mcp.json{
"mcpServers": {
"voicemode": {
"command": "uvx",
"args": [
"voice-mode==9.0.1"
]
}
}
}Runs on your machine.
claude mcp add voicemode -- uvx voice-mode==9.0.1
Runs on your machine.
codex mcp add voicemode -- uvx voice-mode==9.0.1
Runs on your machine.
~/.gemini/config/mcp_config.json{
"mcpServers": {
"voicemode": {
"command": "uvx",
"args": [
"voice-mode==9.0.1"
]
}
}
}Runs on your machine.
gemini mcp add voicemode uvx voice-mode==9.0.1
Runs on your machine.
Runs on your machine.
opencode mcp add voicemode -- uvx voice-mode==9.0.1
Runs on your machine.
~/.hermes/config.yaml{
"mcp_servers": {
"voicemode": {
"command": "uvx",
"args": [
"voice-mode==9.0.1"
]
}
}
}OPENAI_API_KEYOptionalSecretOpenAI API key for cloud-based STT/TTS (optional - local services can be installed)VOICEMODE_DEBUGOptionalEnable debug mode with detailed logging (true/false)VOICEMODE_SKIP_TTSOptionalSkip TTS and show text only for faster response (true/false)VOICEMODE_PREFER_LOCALOptionalPrefer local services over cloud when available (true/false, default: true)VOICEMODE_AUDIO_FORMATOptionalAudio format: pcm, mp3, wav, flac, aac, opus (default: pcm)VOICEMODE_WHISPER_MODELOptionalWhisper model: tiny, base, small, medium, large (default: base)VOICEMODE_DISABLE_SILENCE_DETECTIONOptionalDisable silence detection for continuous recording (true/false)Scanned and passed · checked 5 h ago
Nobody has shared how they use VoiceMode yet.
Security scan
Passed
Safe to install in every agent it supports. Runs on your machine.
Plugins people install alongside VoiceMode.
Ask for a video and your agent writes it as HTML, then HyperFrames renders it to MP4: promos, explainers, captions and motion graphics.
Turns screenshots or a description into a narrated product demo video, with a story arc, designed scenes and an FFmpeg build script.
Lets your agent search screenpipe's local recordings of everything on your screen and microphone, from meetings to what you typed.
