mcp-computer-use
Enterprise-grade macOS MCP server that gives AI agents eyes, hands, and a terminal, enabling screen capture, mouse/keyboard control, shell commands, and file operations.
README
<<<<<<< HEAD
mcp-computer-use
Enterprise-grade macOS MCP server that gives your AI agent eyes, hands, and a terminal.
Tools
-
Screenshot / display
screenshot— capture a display and return a base64 PNG.screenshot_region— capture a region of the screen.get_display_info— list connected displays.get_cursor_position— get the mouse pointer location.ocr_screenshot— run OCR on the screen and return all text.find_text_on_screen— find bounding boxes of text on the screen.
-
Mouse / keyboard
mouse_move,mouse_click,mouse_scrollkeyboard_type,key,hold_key
-
System / apps
open_app— open/activate an app by name.list_windows— list visible windows.focus_window— focus a window.clipboard_get,clipboard_setrun_shell_command— run a shell command (allowlisted).confirm_sensitive_action— approve a pending shell or file action.get_status— show server status and permission state.stop— stop the MCP server process.
-
File system
read_file,write_file,list_dir,delete_file
-
OCR / automation
ocr_screenshot,find_text_on_screenclick_text— click text found by OCR.
-
Processes
process_start— start a long-running shell command.process_read— read output from a running process.process_kill— send a signal to a running process.
-
Batch
batch_operations— run a JSON list of operations sequentially.
Setup
cd /Users/curnutte/CascadeProjects/mcp-computer-use
python3.11 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
Permissions
Grant your terminal/IDE the following permissions in System Settings → Privacy & Security:
- Screen Recording — needed for screenshots, OCR, and Quartz window listing.
- Accessibility — needed for mouse/keyboard control and window focus.
- Input Monitoring — needed for the
Ctrl+Alt+Qglobal kill switch (pynput will warn until this is granted).
If the pynput warning "This process is not trusted! Input event monitoring will not be possible..." appears, the kill switch is armed but cannot receive hotkeys until the host process is added to Input Monitoring.
Test
source .venv/bin/activate
python test_client.py
Windsurf config
Add to ~/.codeium/windsurf/mcp_config.json:
{
"mcpServers": {
"mcp-computer-use": {
"command": "/Users/curnutte/CascadeProjects/mcp-computer-use/.venv/bin/python",
"args": ["-m", "mcp_computer_use"],
"cwd": "/Users/curnutte/CascadeProjects/mcp-computer-use"
}
}
}
Restart Windsurf.
Using the skill
The use_skill.py script demonstrates how the server can be used to run terminal commands and manage the project.
source .venv/bin/activate
python use_skill.py
Configuration
Create ~/.mcp-computer-use/config.json to override defaults:
{
"max_screenshot_dim": 1280,
"allowed_shell_commands": ["git", "python", "python3", "node", "npm", "ls", "pwd", "cat", "echo", "which"],
"blocked_shell_commands": ["rm -rf", "sudo", "mkfs", "dd", ">/dev/null", "shutdown", "reboot", "poweroff"],
"confirm_sensitive": true
}
Kill switch
The server arms a global Ctrl+Alt+Q hotkey via pynput when supported. If it cannot be armed, the server still runs and the process can be killed by the user or an agent calling the stop tool.
Menu-bar app (experimental)
A native macOS menu-bar wrapper lives in MacMenuBar/. It keeps mcp-computer-use
running in the top-right status bar, so the server works from any IDE or CLI:
cd MacMenuBar/MCPMenuBar
./build_app.sh
./install.sh
Then load the LaunchAgent and point your IDE MCP config at
MacMenuBar/bridge/mcp_bridge.py. See MacMenuBar/README.md and
MacMenuBar/integration.md for full details.
License
MIT
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.