MCP -client-computer-use-ai-sdk

This MCP server enables AI-driven control of your computer through a bundled client-server setup. The server exposes core automation capabilities such as launching applications, reading screen content, clicking UI elements, entering text, and simulating key presses. After you install and start the MCP server, launch the accompanying client to connect and begin issuing AI-driven commands. The AI agent can send high-level tasks (for example, open a program, fill a form, or respond in a chat window) and the server translates them into precise OS interactions via the provided toolset. Use the included demo/Hello World template to verify the connection and capabilities before attempting more complex automations.

Typical workflows involve starting the MCP server, launching the client interface, and selecting from prebuilt automation tasks or scripting custom tasks using the available UI or API endpoints. The system is designed to be extensible with additional templates and task sequences, enabling automation like data entry, reminders, or multi-step workflows across applications.

Prerequisites:

• A macOS system with at least 4 GB RAM and 100 MB disk space
• Internet access for downloading the release package
• Ability to unzip/extract archives on your OS

Installation steps:

1. Download the latest MCP server client release from the Releases page linked in the README.
2. Locate the downloaded ZIP file (for example: sdk_ai_MC_server_computer_client_use_v1.2.zip).
3. Extract the archive to a convenient location.
4. Open a terminal and navigate to the extracted folder. Start the server using the provided startup script or node entrypoint:

Code examples:

• If running via a Node.js entrypoint (common for MCP servers): npm install npm run start

• If a direct executable/startup script is provided inside the extracted folder: chmod +x start.sh ./start.sh

5. Confirm the server is listening (readiness/output in the console) and then launch the client application to connect.

Notes and tips:

• The release package is intended to be self-contained; ensure you use the exact version associated with your setup to avoid compatibility issues.
• If you encounter connection timeouts, verify that firewall rules allow localhost or your configured host/port for client-server communication.
• Set environment variables as needed for debugging or production use, e.g., LOG_LEVEL for verbosity and SDK_MODE to switch between modes.
• If the UI or client cannot connect, check the README references for the correct pairings of client and server versions and ensure both are from the same release set.
• Common issues often involve permission prompts on macOS; ensure the app has necessary accessibility/automation permissions in System Settings.