返回 MCP 目录
public公开dns本地运行

mcp服务提醒

一种模型上下文协议服务,当AI代理任务完成时发送桌面通知和警报声音,与Claude Desktop和Cursor等各种LLM客户端集成。

article

README

MseeP.ai Security Assessment Badge PyPI version

📢 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!

help

运行方式说明

cloud

托管运行

托管运行通常表示这个 MCP Server 由服务方环境承载,用户一般按页面提供的连接方式或授权流程接入,不需要在本地长期启动一个 MCP 进程

  1. 打开服务方连接页
  2. 完成授权或复制端点
  3. 在 MCP 客户端中连接
terminal

本地运行 / 其它方式

本地运行通常需要用户在自己的电脑或服务器上安装依赖,把 server_config 复制到 MCP 客户端,并按 env_schema 补齐环境变量、密钥或其它配置

  1. 复制 server_config
  2. 安装所需依赖
  3. 补齐环境变量后重启客户端