Sound Notification MCP

Sound Notification MCP (JPN followed By ENG)

This project is an MCP server that lets AI coding assistants like Windsurf and Cursor play notification sounds when they need user attention. It can play notification sounds when coding is finished or when user approval is needed.

Notes & Disclaimer

• *When adding settings, be sure to add the setting to the rules (memory for Windsurf) as a set.
• Currently it only supports MacOS.
• Since the MCP CALL is left to the discretion of the LLM, there is a possibility that sounds may be heard at unexpected times or that the device may not function as expected.
• In both Cursor and Windsurf, there is often no confirmation prompt before executing a command.
• Windsurf may ring the notification sound multiple times when you exit the app. (Please note that this will consume credits.)

set up

Prerequisites

• Node.js (v18 or higher)
• Yarn

install

Now you're ready to go.

Settings for Windsurf and Cursor

Cursor

A sample is listed in /.cursor/mcp.json. (If you write it in /.cursor/mcp.json, you will need to write it for each project. If you want to enable it globally, please set it from the settings icon.)

Windsurf

Write the settings in mcp_config.json . It is fine to use the same contents as /.cursor/mcp.json.

Customization

Change notification sound

Change the following in src/index.ts file:

The system sounds available on macOS are located in /System/Library/Sounds/ directory:

• Bottle.aiff
• Frog.aiff
• Funk.aiff
• Glass.aiff
• Hero.aiff
• Morse.aiff
• Ping.aiff
• Pop.aiff
• Purr.aiff
• Sosumi.aiff
• Submarine.aiff
• Tink.aiff

troubleshooting

If there is no sound

1. Make sure your system volume is set properly.
2. Please check that the specified audio file exists.
3. Check the terminal output for error messages.

If the MCP server does not start

1. Make sure you have Node.js and Yarn installed correctly.
2. Make sure the dependencies are installed correctly:
   yarn install
3. Check for build errors:
   yarn build

license

MIT

English Version

This project is an MCP server that plays notification sounds when AI coding assistants like Windsurf or Cursor require user attention. It can play notification sounds when coding is complete or when user approval is needed.

Notes & Disclaimer

• * When adding settings, please always add corresponding entries to the rules.
• Currently only supports MAC as the creator uses a Mac and cannot verify on Windows.
• Since MCP CALLS depend on LLM judgment, sounds may play at unexpected times or not work as expected.
• For Cursor, sounds may sometimes not play before command execution approval.
• With Windsurf, there is a phenomenon where notification sounds may play multiple times.

Setup

Prerequisites

• Node.js (v18 or higher)
• Yarn

Installation

Now you're ready to go.

Configuration for Windsurf and Cursor

Cursor

A sample configuration is provided in /.cursor/mcp.json (Writing to /.cursor/mcp.json requires configuration for each project. If you want to enable it globally, please configure it from the settings icon.)

Windsurf

Add configuration to mcp_config.json . The content can be exactly the same as /.cursor/mcp.json .

For Windsurf, it is also strongly recommended to:

1. Add the notification instructions to your global_rules.md file
2. Add the notification settings to your memories

This ensures that the AI assistant consistently uses sound notifications when needed.

Customization

Changing the Notification Sound

Modify the following section in the src/index.ts file:

System sounds available on macOS can be found in the /System/Library/Sounds/ directory:

• Bottle.aiff
• Frog.aiff
• Funk.aiff
• Glass.aiff
• Hero.aiff
• Morse.aiff
• Ping.aiff
• Pop.aiff
• Purr.aiff
• Sosumi.aiff
• Submarine.aiff
• Tink.aiff

Troubleshooting

If No Sound Plays

1. Check that the system volume is properly set.
2. Verify that the specified sound file exists.
3. Check the terminal output for error messages.

If the MCP Server Fails to Start

1. Verify that Node.js and Yarn are correctly installed.
2. Check that dependencies are properly installed:
   yarn install
3. Check for build errors:
   yarn build

License

MIT