VoiceMode MCP Server: Voice Conversations for Claude Code and Other Apps

• Speaks to you and listens for a reply

  Says a message out loud through your speakers and, by default, listens through your microphone for your answer, which comes back as text. It can also narrate without listening, or ask several questions in a row.

  Try: “Ask me out loud which file I want to work on first.”

  Write · converse

• Steps away without losing the floor

  Pauses the voice conversation for a set time while keeping the speaking slot, so another agent cannot jump in before the assistant comes back.

  Try: “Pause the conversation for two minutes while you run the tests.”

  Write · pause_conversation

• Manages who may speak

  Shows or changes the single-speaker lock that stops several agents talking at once. Actions include status, queue, wait, heartbeat, leave, give, bump and release.

  Try: “Show who holds the voice channel right now.”

  Write · conch

• Lists microphones and speakers

  Lists the audio input and output devices on your computer.

  Try: “What audio devices does VoiceMode see?”

  Read · check_audio_devices

• Checks audio packages

  Checks the system audio packages, PulseAudio state and Python audio libraries, and gives install commands for anything missing.

  Try: “Check whether my audio setup is complete.”

  Read · check_audio_dependencies

• Checks voice services

  Checks speech-to-text, text-to-speech, LiveKit and audio devices. The tool says to use it for debugging only, since services fail over on their own.

  Try: “Voice stopped working. Check the voice services.”

  Network · voice_status

• Lists available voices

  Lists the text-to-speech voices for each provider, such as OpenAI or Kokoro.

  Try: “Which Kokoro voices can I use?”

  Network · list_tts_voices

• Shows version and provider info

  Returns the installed version, configuration and provider status, to help debug issues.

  Try: “Show the VoiceMode diagnostic info.”

  Read · voice_mode_info

• Re-checks voice endpoints

  Runs fresh health checks on speech endpoints after you start or stop a service, without restarting the server.

  Try: “Refresh the text-to-speech providers.”

  Network · refresh_provider_registry

• Shows one voice endpoint

  Returns details of one provider endpoint by its base URL, including its models and voices.

  Try: “Show details for http://127.0.0.1:8880/v1.”

  Network · get_provider_details

• Shows all discovered voice endpoints

  Returns the registry of speech endpoints found, with models, voices, provider type and last check time.

  Try: “Show the voice provider registry.”

  Read · voice_registry

• Lists the VOICEMODE_ settings

  Lists every VOICEMODE_ setting and what it does.

  Try: “What VoiceMode settings can I change?”

  Read · list_config_keys

• Shows which config files are used

  Shows the global ~/.voicemode/voicemode.env file and project .voicemode.env files that are loaded, in order.

  Try: “Which VoiceMode config files are active?”

  Files · show_config_files

• Reloads settings

  Reloads VoiceMode settings from the global and project env files and the environment, and clears caches.

  Try: “Reload the VoiceMode configuration.”

  Files · config_reload

• Changes a setting in voicemode.env

  Writes a new value for a VOICEMODE_ setting into your voicemode.env file.

  Try: “Set VOICEMODE_VOICES to af_sky.”

  Write · update_config

• Shows the live performance dashboard

  Shows response counts, success rate and average speech and transcription timings for this session.

  Try: “Show voice conversation stats.”

  Read · voice_statistics

• Gives a short stats summary

  Returns session length, interaction count and average turnaround and time-to-first-audio.

  Try: “Give me a one-paragraph voice performance summary.”

  Read · voice_statistics_summary

• Lists recent interactions

  Shows the most recent voice interactions with timing, up to 50.

  Try: “Show my last 10 voice interactions.”

  Read · voice_statistics_recent

• Exports stats as JSON

  Returns all tracked metrics and calculated statistics as JSON.

  Try: “Export the voice stats as JSON.”

  Read · voice_statistics_export

• Clears the stats

  Clears all tracked voice metrics and starts a new session timer.

  Try: “Reset the voice statistics.”

  Write · voice_statistics_reset

• Starts, stops and checks voice services

  Manages the Whisper, Kokoro and VoiceMode services. Actions are status, start, stop, restart, enable and disable at login, and logs.

  Try: “Start the Kokoro service and enable it at login.”

  Runs code · service

• Installs whisper.cpp speech-to-text

  Installs whisper.cpp on your computer with a speech model, and can enable it as a service. It supports macOS and Linux and may use your GPU.

  Try: “Install Whisper with the base model.”

  Runs code · whisper_install

• Downloads Whisper models

  Downloads Whisper model files into the configured folder, with pre-built Core ML versions on Apple Silicon.

  Try: “Download the small Whisper model.”

  Write · whisper_model_install

• Removes whisper.cpp

  Stops the Whisper service, removes its service setup and install, and can also delete downloaded models and data.

  Try: “Uninstall Whisper but keep the models.”

  Delete · whisper_uninstall

• Installs Kokoro text-to-speech

  Clones the Kokoro-FastAPI project into ~/.voicemode/services/kokoro, uses its start script and installs a login service.

  Try: “Install Kokoro for local text-to-speech.”

  Runs code · kokoro_install

• Removes Kokoro

  Stops the Kokoro service, removes its service setup and install, and can also remove models and data.

  Try: “Uninstall Kokoro and remove its models.”

  Delete · kokoro_uninstall

Watch out for

• It can install, run and remove software

  whisper_install, kokoro_install and service download code, set up background services and start them at login. The two uninstall tools can also delete models and data. github.com

• It listens through your microphone

  converse records your speech. With OpenAI selected, the audio goes to OpenAI. With local Whisper it stays on your computer. github.com

• It can edit its own settings file

  update_config writes values into your voicemode.env file, which controls voices, services and keys. github.com

• You can load fewer tools

  The code reads VOICEMODE_TOOLS_ENABLED to load only a list of tools, or VOICEMODE_TOOLS_DISABLED to leave some out. github.com

Security advisories