Voice Recorder

An MCP server that records audio from a microphone and transcribes it with Whisper.

python

6

GitHub Stars

python

Language

4 months ago

First Indexed

3 weeks ago

Catalog Refreshed

Documentation & install

Readme and setup notes from the catalogue, plus a client-ready config you can copy for your MCP host.

Installation

Add the following to your MCP client configuration file.

Configuration

View docs

{
  "mcpServers": {
    "defibax-mcp_servers": {
      "command": "voice-recorder-mcp",
      "args": [],
      "env": {
        "SAMPLE_RATE": "44100",
        "MAX_DURATION": "120",
        "WHISPER_MODEL": "small.en"
      }
    }
  }
}

You can run a microphone-based MCP server that records audio and transcribes it using Whisper. This setup is practical for quick voice notes, meeting transcripts, or integrating with Goose as a custom extension to trigger actions from transcriptions.

How to use

To run the server locally, you start the voice recorder MCP executable. By default it uses the base.en Whisper model for fast, reliable English transcription.

You can start a transcription session by simply launching the server, then say something into your microphone. Use the available tools to control recording and transcription. You can also run a specific model or adjust the sample rate if your environment requires it.

If you want to test the server with the MCP Inspector, install the inspector tool and connect to your running server to interactively send transcription results and verify behavior.

When you want Goose to act on transcriptions, configure Goose to invoke the server executable and, if desired, specify a model. For example, you can set a command that runs the server with a chosen model and then instruct Goose to respond to transcriptions by performing actions requested in the text.

How to install

Prerequisites: Python 3.12 or newer and an audio input device (microphone). You may also install a command-line tool for testing MCP endpoints if you want an interactive workflow.

# Install from source
git clone https://github.com/DefiBax/voice-recorder-mcp.git
cd voice-recorder-mcp
pip install -e .

Run the server with default settings to start using the base English Whisper model. You can also choose a different model or adjust the sample rate as needed.

voice-recorder-mcp

voice-recorder-mcp --model medium.en

voice-recorder-mcp --sample-rate 44100

To test the server with the MCP Inspector tool, install the inspector and connect to your running server to validate behavior.

# Install the MCP Inspector
npm install -g @modelcontextprotocol/inspector

# Run your server with the inspector
npx @modelcontextprotocol/inspector voice-recorder-mcp

Configuration and usage notes

Set runtime options using environment variables to customize behavior. For example, set the Whisper model, sample rate, and maximum recording duration before starting the server.

export WHISPER_MODEL=small.en
export SAMPLE_RATE=44100
export MAX_DURATION=120

voice-recorder-mcp

Security and integration tips

Ensure microphone permissions are granted on your operating system. When integrating with Goose, verify the command path is correct and that the extension has access to the server executable. If you adjust models or rates, validate the transcription quality and consider lowering the sample rate if you encounter audio quality issues.

Available tools

start_recording

Start recording audio from the default microphone.

stop_and_transcribe

Stop the recording and transcribe the captured audio to text.

record_and_transcribe

Record audio for a specified duration and transcribe it to text.