Overview
Computer Control MCP is a Python-based MCP server that exposes a suite of actions to remotely control a computer's user interface. It leverages PyAutoGUI for mouse and keyboard interactions and integrates OCR via RapidOCR (ONNXRuntime) for text extraction from screenshots, all while aiming for zero external dependencies beyond these libraries. The server provides functions to move and click the mouse, perform drag-and-drop, type text at the current cursor position, press and hold keys, and execute multi-key sequences. It supports screen capture of the entire display or specific windows, with optional saving to a downloads folder, and an OCR-enabled variant that returns text coordinates. Additional utilities include listing open windows, activating a window by title or pattern, fetching the current screen size, and waiting for a specified duration in milliseconds. Deployment is described via uvx with a sample mcpServers config, or via a standard pip install, and the repository includes development setup, test instructions, and an API reference. The project emphasizes ease of use and zero-external-dependency packaging.
Features
Control mouse movements and clicks
Move the cursor, click at coordinates, and perform drag-and-drop with configurable buttons and durations.
Type text at the current cursor position
Type text at the active cursor, including typing, key down/up, and sequences.
Take screenshots of the entire screen or specific windows with optional saving to downloads directory
Capture images of the full screen or a target window, with optional save to a downloads folder.
Extract text from screenshots using OCR
Run OCR on captured images to extract text and coordinates.
List and activate windows
Enumerate open windows and bring a chosen window to the foreground by title or pattern.
Press keyboard keys
Press single keys, key sequences, and multi-key combinations.
Drag and drop operations
Perform drag-and-drop actions between coordinates with optional duration.
