computer-split-screen-mcp
A high-performance, cross-platform MCP server for split-screen window management on Windows and macOS, providing 16 window manipulation tools.
README
Computer Split Screen MCP
A high-performance, cross-platform Model Context Protocol (MCP) server that provides reliable split-screen window management for both Windows and macOS. This server exposes 16 window manipulation tools through the MCP protocol, enabling AI assistants and other MCP clients to control desktop window layouts with precision.
🚀 Features
- Cross-platform support: Windows and macOS with native optimizations
- 16 window management tools: Comprehensive split-screen layouts and controls
- High performance: Windows: 15-60ms, macOS: 50-150ms per operation
- Reliability first: Robust fallback mechanisms and error handling
- MCP integration: Full Model Context Protocol server support via stdio transport
- Smart focus detection: Advanced window focus detection with platform-specific optimizations
🛠️ Installation
Prerequisites
- Python: 3.9 or higher
- Package Manager:
uvx(recommended) orpip
Install via uvx (Recommended)
uvx install computer-split-screen-mcp
Install via pip
pip install computer-split-screen-mcp
🔧 MCP Client Configuration
Configure your MCP client with the following settings:
{
"mcpServers": {
"computer-split-screen": {
"command": "uvx",
"args": ["computer-split-screen-mcp"],
"env": {}
}
}
}
Alternative Configuration (if not using uvx)
{
"mcpServers": {
"computer-split-screen": {
"command": "python",
"args": ["-m", "computer-split-screen-mcp"],
"env": {}
}
}
}
🎯 Available Tools
Split-Screen Layouts
Halves (2-way splits)
left-half-screen- Snap current window to left halfright-half-screen- Snap current window to right halftop-half-screen- Snap current window to top halfbottom-half-screen- Snap current window to bottom half
Quadrants (4-way splits)
top-left-screen- Top-left quadrant (1/4 screen)top-right-screen- Top-right quadrant (1/4 screen)bottom-left-screen- Bottom-left quadrant (1/4 screen)bottom-right-screen- Bottom-right quadrant (1/4 screen)
Thirds (3-way splits)
left-one-third-screen- Left third (1/3 screen)middle-one-third-screen- Middle third (1/3 screen)right-one-third-screen- Right third (1/3 screen)
Two-Thirds (2/3 splits)
left-two-thirds-screen- Left two-thirds (2/3 screen)right-two-thirds-screen- Right two-thirds (2/3 screen)
Window Controls
maximize-screen- OS maximize (bordered, taskbar visible)fullscreen-screen- Fullscreen mode (platform-specific behavior)minimize-screen- Minimize window to taskbar/dock
⚡ Performance Characteristics
Windows Performance
- Total time: 15-60ms per operation
- Detection: 2-5ms (direct Win32 API)
- Manipulation: 10-47ms (SetWindowPos with corrections)
- Best case: 15-25ms for simple splits
- Typical case: 20-35ms for most operations
macOS Performance
- Total time: 50-150ms per operation
- Detection: 20-50ms (AppleScript + Accessibility API)
- Manipulation: 15-55ms (AXUIElement operations)
- Best case: 50-80ms for simple splits
- Typical case: 80-120ms for most operations
🔍 Technical Architecture
Windows Implementation
- Direct Win32 API calls via
pywin32 - DWM integration for accurate frame bounds
- DPI-aware positioning for high-resolution displays
- Two-pass positioning for precise window placement
macOS Implementation
- AppleScript focus detection for reliability
- Accessibility API for window manipulation
- Fallback mechanisms for edge cases
- Screen-aware positioning respecting menu bar and Dock
Cross-Platform Features
- Automatic platform detection
- Conditional dependency loading
- Unified API interface
- Error handling and recovery
📋 Platform Dependencies
Windows
- Required:
pywin32>=306 - Purpose: Win32 API access, DWM integration, window manipulation
macOS
- Required:
pyobjc-core>=10.1,<11,pyobjc-framework-Cocoa>=10.1,<11,pyobjc-framework-Quartz>=10.1,<11,pyobjc-framework-ApplicationServices>=10.1,<11 - Purpose: Accessibility API, AppleScript integration, window management
🚨 Troubleshooting
Common Issues
macOS Focus Detection Problems
- Symptom: Windows don't move or wrong windows are selected
- Solution: Ensure Terminal has Accessibility permissions in System Preferences > Security & Privacy > Privacy > Accessibility
Windows DPI Issues
- Symptom: Windows positioned incorrectly on high-DPI displays
- Solution: The server automatically handles DPI awareness, but ensure
pywin32>=306is installed
MCP Client Connection Issues
- Symptom: Functions work when called directly but fail via MCP
- Solution: Check MCP client logs, ensure proper configuration, restart MCP client
Performance Optimization
- First run: May be slower due to system warm-up
- Subsequent runs: Should be consistently fast
- Complex apps: Safari, Chrome may take longer due to window structure
🔧 Development
Project Structure
computer-split-screen/
├── src/splitscreen_mcp/
│ ├── __init__.py # Package initialization
│ ├── __main__.py # MCP server entry point
│ └── window_actions.py # Core window management logic
├── pyproject.toml # Project configuration
├── README.md # This file
└── LICENSE # MIT License
Building from Source
git clone https://github.com/Beta0415/computer-split-screen-mcp.git
cd computer-split-screen-mcp
uvx install -e .
Running Tests
# Test window detection
python3 -c "from src.splitscreen_mcp.window_actions import left_half_window; left_half_window()"
# Test MCP server
uvx run computer-split-screen-mcp
📄 License
This project is licensed under the MIT License - see the LICENSE file for details.
📊 Version History
-
v1.5.0 - Current stable release
- Cross-platform window management
- 16 comprehensive tools
- High-performance implementation
- Full MCP protocol support
-
v1.1.0 - v.1.4.5 - In-progress testing & updates
-
v.1.2.0 - Outdated release
- macOS uses pure AppleScript window relocation method
-
v1.0.0 - Intitial release
Built with ❤️ for the MCP community
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。