Desktop Commander MCP Server

A comprehensive Model Context Protocol (MCP) server that gives AI assistants like Claude full control over your desktop — monitor system resources, manage windows, capture screenshots, control the clipboard, launch applications, and more.

23 tools across 4 categories, turning Claude into a true desktop assistant.

Why Desktop Commander?

Claude is powerful, but blind and handless on your desktop. Desktop Commander fixes that:

• "What's eating my CPU?" — instantly see top resource consumers

• "Take a screenshot" — Claude can actually see your screen

• "Arrange my windows side by side" — window management by natural language

• "Copy this to clipboard" — seamless data transfer

• "Is port 3000 in use?" — debug development issues

• "Open VS Code in my project folder" — launch apps with context

Related MCP server: openowl

Tools

System Monitoring (10 tools)

Window Management (5 tools)

Desktop Control (5 tools)

Application Management (3 tools)

Installation

git clone https://github.com/19miha99/system-monitor-mcp.git
cd system-monitor-mcp
pip install -e .

Usage with Claude Desktop

Add to your Claude Desktop config (claude_desktop_config.json):

{
  "mcpServers": {
    "desktop-commander": {
      "command": "python",
      "args": ["-m", "system_monitor_mcp.server"]
    }
  }
}

Usage with Claude Code

claude mcp add desktop-commander -- python -m system_monitor_mcp.server

Example Workflows

System Health Check

"Give me a complete system health check"

Claude calls get_system_overview, get_top_consumers, get_network_stats and gives you a comprehensive report.

Debug a Slow Computer

"My computer is slow, what's going on?"

Claude checks get_top_consumers for CPU and memory, identifies the culprit, and offers to kill_process if needed.

Window Arrangement

"Put Chrome on the left half and VS Code on the right half of my screen"

Claude uses list_windows to find them, get_display_info to know the screen size, then arrange_window to position each one.

Port Conflict Resolution

"Something is using port 8080, find and kill it"

Claude calls find_process_by_port(8080), shows you what's there, and kill_process to free the port.

Visual Debugging

"Take a screenshot and tell me what you see"

Claude uses capture_screenshot, then reads the image to analyze what's on screen.

App Launcher

"Open my project in VS Code"

Claude calls launch_application("code", args=["C:/Users/you/project"]).

Architecture

src/system_monitor_mcp/
├── app.py          # FastMCP instance (shared across modules)
├── server.py       # Entry point — imports all tool modules
├── helpers.py      # Shared utilities (format_bytes, format_uptime)
├── monitor.py      # 10 system monitoring tools
├── windows.py      # 5 window management tools
├── desktop.py      # 5 desktop control tools (screenshots, clipboard, etc.)
└── apps.py         # 3 application management tools

Requirements

• Python 3.10+

• psutil — cross-platform system monitoring

• mcp — Model Context Protocol SDK

• Pillow — screenshot capture

• pywin32 — Windows API access (auto-installed on Windows)

Platform Support

Primary platform: Windows. Cross-platform support for desktop features is planned.

Safety

• System-critical processes (PID 0, 1, 4) cannot be killed

• The server refuses to kill its own process

• Windows are closed gracefully via WM_CLOSE (not force-terminated)

• All operations provide clear error messages on failure

License

MIT