MCPControl

MCPControl is an MCP server that provides programmatic control over the Windows operating system for AI models. It connects client interfaces like Claude to native Windows desktop environments, allowing agents to manipulate the mouse, send keyboard inputs, manage open windows, read and update clipboard contents, and capture screen states. Software developers, automated QA engineers, and workflow automation builders use this tool to build autonomous agents capable of interacting with GUI-based desktop software. By exposing native Windows API bindings and automation backends like keysender, PowerShell, and AutoHotkey v2, MCPControl enables AI models to visually inspect applications and execute multi-step user workflows directly inside a Windows session. Because the server grants direct input simulation and operating system access, it is designed for controlled virtual machine environments to facilitate UI testing, automated form handling, and desktop interaction research without requiring custom application APIs.

Category: Cloud & Infrastructure

Tags: automation, gui, keyboard, mouse, windows

Visit MCPControl

How to install and configure MCPControl

Follow these steps to install and run MCPControl on Windows: 1. Install required prerequisites using PowerShell as Administrator: powershell winget install Microsoft.VisualStudio.2022.BuildTools --override "--wait --passive --add Microsoft.VisualStudio.Workload.VCTools --includeRecommended" winget install Python.Python.3.12 winget install OpenJS.NodeJS 2. Install MCPControl globally: powershell npm install -g mcp-control 3. Start the server using SSE transport (recommended inside a 1280x720 VM): bash mcp-control --sse 4. Add the server configuration to your Claude configuration file: json { "mcpServers": { "MCPControl": { "transport": "sse", "url": "http://192.168.1.100:3232/mcp" } } } Replace 192.168.1.100:3232 with your actual host IP address and port.

What you can do with MCPControl

• Delegate repetitive user interface testing across desktop applications by having an AI navigate menus, trigger actions, and report issues visually. - Automate supervised data entry and form filling by enabling an agent to switch windows, focus fields, and type inputs. - Conduct AI experimentation and visual reasoning tasks, allowing models to observe screen captures and solve visual interface puzzles. - Build cross-application workflows that bridge legacy Windows software lacking direct API access through screen inspection and simulated input.

Key facts

• https://github.com/Cheffromspace/MCPControl
• Cloud & Infrastructure, Design, Media & Creative
• automation, gui, keyboard, mouse, windows

Part of MCP Servers

How do I install MCPControl?

To install MCPControl, first install Visual Studio 2022 Build Tools with VC++ tools, Python 3.12, and Node.js using winget. Once the build dependencies are ready, run npm install -g mcp-control in your terminal. You can then run the server via the command mcp-control --sse.

What can MCPControl do?

MCPControl enables AI models to control a Windows desktop environment. It provides functionality for moving the mouse, clicking, dragging, sending keyboard keystrokes and text, managing active windows, inspecting screen dimensions, capturing desktop screenshots, and manipulating clipboard data through modular automation backends like keysender, PowerShell, or AutoHotkey.

Which MCP clients work with MCPControl?

MCPControl works with any Model Context Protocol client supporting Server-Sent Events (SSE) network transport, such as Claude Desktop. The client connects over HTTP or HTTPS to the server endpoint hosted on a local machine or a dedicated Windows virtual machine.

What operating systems are supported by MCPControl?

MCPControl currently supports only Windows operating systems. The server depends on native Windows modules such as keysender and Windows APIs to execute input commands and desktop window management operations.

Is MCPControl open source?

Yes, MCPControl is open source software published under the MIT License. The code and issue tracker are publicly hosted on GitHub.