📢 MCP Notify Server

English | 中文

A MCP server that send desktop notifications with sound effect when agent tasks are completed.

🥩 Features

• Send system desktop notifications after agent tasks completion
• Play alert sounds to grab user attention, with sound file inside.
• Cross-platform support (Windows, macOS, Linux)
• Based on standard MCP protocol, integrates with various LLM clients

⏬ Installation

Install using uv package manager

git clone https://github.com/Cactusinhand/mcp_server_notify.git
cd mcp_server_notify

uv venv
source .venv/Scripts/activate

uv pip install mcp-server-notify
# or
pip install mcp-server-notify

After installation, call the module directly to check if installation was successful:

python -m mcp_server_notify

This module accepts --debug or --file option, we can use it like:

python -m mcp_server_notify --debug
python -m mcp_server_notify --debug --log-file=path/to/logfile.log

⚠️❕ Special requirements

** We use Apprise API for our Desktop notification deliver，so we need to install some special requirements in our Desktop **

Windows

# windows:// minimum requirements
pip install pywin32

macOS

# Make sure terminal-notifier is installed into your system
brew install terminal-notifier

📚 Usage

Using with Claude Desktop:

Find the configuration file claude_desktop_config.json

{
    "mcpServers": {
        "NotificationServer": {
            "command": "uv",
            "args": [
              "--directory",
              "path/to/your/mcp_server_notify project",
              "run",
              "mcp-server-notify",
            ]
        }
    }
}

If installed globally, you can also use the python command:

{
    "mcpServers": {
        "NotificationServer": {
            "command": "python",
            "args": [
              "-m",
              "mcp_server_notify",
            ]
        }
    }
}

⚡️ Using with Cursor:

Find the configuration file ~/.cursor/mcp.json or your_project/.cursor/mcp.json

{
    "mcpServers": {
        "NotificationServer": {
            "command": "uv",
            "args": [
              "--directory",
              "path/to/your/mcp_server_notify project",
              "run",
              "mcp-server-notify",
            ]
        }
    }
}

After configuration, simply add a prompt like finally, send me a notification when task finished. at the end of your task input to the AI to trigger notifications.

In Cursor, you can add this prompt as a rule in Cursor Settings -> Rules so you don't have to type it manually each time.

⚡️ Using with VSCode + Copilot:

1. Install the service manager uv/uvx: pip install uv

2. Add the service to VSCode settings:

   Windows %APPDATA%\Code\User\settings.json
   macOS $HOME/Library/Application\ Support/Code/User/settings.json
   Linux $HOME/.config/Code/User/settings.json

   "mcp": {
       "servers": {
           "notifier": {
               "command": "uvx",
               "args": [
                   "mcp-server-notify"
               ],
               "env": {}
           }
       }
   }

3. Make sure you are using the latest VSCode version — it automatically runs MCP services

4. Open VSCode → enable Copilot → switch to agent mode.

5. Type # → you will see the #send_notification option.

6. Ask the agent: run #send_notification (it will handle the notification automatically).

7. Now the Copilot in agent mode can send desktop notifications.

🐳 Running with Docker

Currently not available due to environment compatibility issues. If Docker containers need to trigger host notifications regardless of whether the host OS is Windows, macOS, or Linux, the solution becomes much more complex, and direct use of native notifications is usually not feasible.

Main issues:

1. OS-specific notification systems Each operating system (Windows, macOS, Linux) has its unique notification mechanism.

2. Docker isolation The isolation of Docker containers limits their ability to access host operating system resources directly.

3. Dependency management Need to handle different notification libraries and dependencies for each operating system.

🧾 License

MIT

💻 Contributions

Issues and pull requests are welcome!