Screen Control — Claude Code MCP server: what it does, how to install
Screen Control is a Claude Code MCP server providing desktop screen capture, mouse and keyboard automation. Runs whenever an agent needs direct interaction with local system windows.
MIT ★ 10 updated 2026-10-29
Open on GitHubWhat Screen Control does
Captures screen state via optical character recognition or video streaming and executes synthetic keyboard and mouse inputs. The server processes everything locally on the host machine without forwarding data to third-party cloud infrastructure. It supports window handling, background message dispatching without stealing focus, and visual diff tracking.
Built-in safety controls prevent runaway execution. The service includes an automatic release watchdog for stuck keys, an emergency stop triggered by moving the cursor to the top-left screen corner, and mandatory token authentication using the X-Auth-Token header.
Execution requires Python on either Windows or Linux running an X11 display server. Wayland sessions are unsupported and will reject server startup.
What it helps with
- capture on-screen text with optical character recognition
- send mouse clicks, drags, and scrolling events to desktop windows
- dispatch Unicode keyboard input and application shortcuts
- manipulate background application windows without stealing focus
Example prompts for the agent
- Inspect the desktop display using OCR to locate the target dialog
- Send a mouse click to the specified button coordinates
- Dispatch keystrokes into the active text editor without taking focus
How to install Screen Control
Install
The MCP registry has no launch command for it: see the repository README
Step by step, with commands for each platform: How to add an MCP server to Claude Code, Codex, Cursor, VS Code, Gemini CLI and more
Questions about Screen Control
How to install Screen Control in Claude Code?
Add the server to your MCP client configuration by pointing to the repository runtime using Python.
What does Screen Control do?
It feeds screen captures and OCR output to the agent while translating tool calls into synthetic mouse and keyboard inputs.
What is required to run Screen Control?
You need Python and a desktop running Windows or Linux with X11. Wayland display servers are not supported.