AlertMCP

An MCP tool that provides desktop notification functionality. You can send notifications from AI client tools like Claude and Cursor.

Features

• Set agent name
• Send desktop notifications
• Schedule notifications at specific times
• Schedule notifications after a delay

Installation

About UV Installation

For installing alertmcp and managing dependencies, we recommend using the high-speed Python package manager "uv". It's more efficient than pip for creating virtual environments and installing packages.

UV Installation Method

You can install uv on Windows (PowerShell) or macOS (bash) with the following command:

pip install uv

After installation, you can check the version with:

uv --version

Installation from GitHub

Clone the repository to your local machine.

Using PowerShell (Windows):

# Clone the repository
git clone https://github.com/kousunh/alertmcp.git

# Navigate to the cloned directory
cd alertmcp

# Create a virtual environment
uv venv

# Install the package
uv pip install .

Using Bash (macOS/Linux):

# Clone the repository
git clone https://github.com/kousunh/alertmcp.git

# Navigate to the cloned directory
cd alertmcp

# Create a virtual environment
uv venv

# Install the package
uv pip install .

# For macOS, additionally install pyobjus (in the virtual environment)
uv pip install pyobjus

OS-specific Notes

• Windows: The notification function uses plyer, which is automatically installed as a dependency.
• macOS: The notification function requires plyer and pyobjus. You need to separately install pyobjus even when using uv.

Setting Up and Using with AI Tools

After cloning the directory locally, set up each AI tool as follows:

Setup in Claude Desktop

1. Open Claude desktop settings
2. Find and edit the settings file:
   • macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
   • Windows: %APPDATA%\\Claude\\claude_desktop_config.json
3. Add the following settings:

{
  "mcpServers": {
    "alert_mcp":{
      "command": "uv",
      "args": [
        "--directory",
        "<Path to your Git cloned folder>\\\\alert_mcp",
        "run",
        "python",
        "-m",
        "alertmcp.server"
      ],
      "alwaysAllow": ["add"],
      "disabled": false
    }
  }
}

Setup in Cursor

1. Open Cursor settings and navigate to the MCP Servers section
2. Click "Add New MCP Server"
3. Enter the following settings:
   • Name: AlertMCP (or any name you prefer)
   • Type: command
   • Command: uv --directory <Path to your Git cloned folder>\\alert_mcp run python -m alertmcp.server

JSON Configuration for Cursor

You can also directly edit the Cursor settings file "mcp.json":

{
  "alert_mcp": {
    "command": "uv",
    "args": [
      "--directory",
      "<Path to your Git cloned folder>\\\\alert_mcp",
      "run",
      "python",
      "-m",
      "alertmcp.server"
    ],
    "alwaysAllow": ["add"],
    "disabled": false
  }
}

Setting Agent Name

You can customize the agent name with:

Please set your name to 'AI Agent Name'

Example Instructions

You can give the following instructions to your MCP-enabled AI assistant:

Please send a notification when you finish your task

Please notify me at 15:30 today to "Check presentation materials"

Please notify me in 5 minutes that "It's time to take a break"

Documentation

For Japanese documentation, see README_JA.md

License

MIT License