MCPControlSAFE
MCP server for Windows OS automation
Overview
From the repository's own README, as read at the audited commit. Badges and raw HTML are left out.
Windows control server for the Model Context Protocol, providing programmatic control over system operations including mouse, keyboard, window management, and screen capture functionality.
Note: This project currently supports Windows only.
🔥 Why MCPControl?
MCPControl bridges the gap between AI models and your desktop, enabling secure, programmatic control of:
- 🖱️ Mouse movements and clicks
- ⌨️ Keyboard input and shortcuts
- 🪟 Window management
- 📸 Screen capture and analysis
- 📋 Clipboard operations
🔌 Quick Start
Prerequisites
- Install Build Tools (including VC++ workload)
# Run as Administrator - may take a few minutes to complete winget install Microsoft.VisualStudio.2022.BuildTools --override "--wait --passive --add Microsoft.VisualStudio.Workload.VCTools --includeRecommended"
- Install Python (if not already installed)
# Install Python (required for node-gyp) winget install Python.Python.3.12
- Install Node.js
# Install latest LTS version winget install OpenJS.NodeJS
Installation
- Install MCPControl Package
npm install -g mcp-control
Configuration
MCPControl works best in a virtual machine at 1280x720 resolution for optimal click accuracy.
Configure your Claude client to connect to MCPControl via SSE transport:
Option 1: Direct SSE Connection
For connecting to an MCPControl server running on a VM or remote machine:
{
"mcpServers": {
"MCPControl"ac93f17d1992OBSERVED · 2026-10-07Connect
Built from this server's own package name, version and transport as found in its source — not copied from anyone's documentation, so it cannot drift against a page we do not control. Replace the environment placeholders with a token scoped to the least it needs.
claude mcp add mcp-control --env AUTOHOTKEY_PATH=${AUTOHOTKEY_PATH} --env AUTOMATION_KEYBOARD_PROVIDER=${AUTOMATION_KEYBOARD_PROVIDER} -- npx -y [email protected]{
"mcpServers": {
"mcp-control": {
"command": "npx",
"args": [
"-y",
"[email protected]"
],
"env": {
"AUTOHOTKEY_PATH": "${AUTOHOTKEY_PATH}",
"AUTOMATION_KEYBOARD_PROVIDER": "${AUTOMATION_KEYBOARD_PROVIDER}"
}
}
}
}Exposed tools (23)
18 read · 4 write · 1 destructive. Blast radius: 1 tool can delete or overwrite — an agent that can be talked into calling a tool can be talked into calling this one.
| Tool | Risk | Description |
|---|---|---|
clear_clipboard | destructive | Clear the clipboard content |
click_at | write | Move mouse to coordinates, click, then return to original position |
click_mouse | read | Click the mouse at the current position |
double_click | read | Double click at current or specified position |
drag_mouse | read | Drag the mouse from one position to another |
focus_window | read | Focus a specific window by its title |
get_active_window | read | Get information about the currently active window |
get_clipboard_content | read | Get the current text content from the clipboard |
get_cursor_position | read | Get the current cursor position |
get_screen_size | read | Get the screen dimensions |
get_screenshot | read | Take a screenshot optimized for AI readability, especially for text-heavy content. Uses default settings: JPEG format, 85% quality, grayscale enabled, and 1280px width (preserving aspect ratio). Supports region capture, format options, quality adjustment, and custom resize settings. |
has_clipboard_text | read | Check if the clipboard contains text |
hold_key | read | Hold or release a keyboard key with optional duration |
minimize_window | read | Minimize a specific window by its title (currently unsupported) |
move_mouse | write | Move the mouse cursor to specific coordinates |
press_key | read | Press a specific keyboard key |
press_key_combination | read | Press multiple keys simultaneously (e.g., keyboard shortcuts) |
reposition_window | write | Move a specific window to new coordinates |
resize_window | read | Resize a specific window by its title |
restore_window | read | Restore a minimized window by its title (currently unsupported) |
scroll_mouse | read | Scroll the mouse wheel up or down |
set_clipboard_content | write | Set text content to the clipboard |
type_text | read | Type text using the keyboard |
Trust audit
SAFEgrade B · trust 89/100 Nothing in the source contradicts what it says it does. Grade A is reserved for packages that have also passed the behavioural sandbox.
| Layer | What it checks | Result |
|---|---|---|
| L0 | Provenance & inventory | PASS |
| L1 | Static analysis of the code | PASS |
| L2 | Instruction surface (what it tells the agent) | PASS |
| L3 | Class-specific surface | WARN |
| L4 | Behavioural (sandbox) | SKIPPED |
What the source does
- Filesystem
- declared (6 observation(s))
- Network
- none-observed
- Shell
- declared (5 observation(s))
- Dependencies
- not all pinned
- Secrets in source
- none-found
Findings (14)
clear_clipboard
.lintstagedrc
.prettierignore
import { WindowsControlResponse } from '../../types/responses.js';import { ClipboardAutomation } from '../../interfaces/automation.js';import { ClipboardInput } from '../../types/common.js';import { AutomationProvider } from '../../interfaces/provider.js';} from '../../interfaces/automation.js';
"url": "http://192.168.1.100:3232/mcp"
"url": "http://192.168.1.100:3232/mcp"
@modelcontextprotocol/sdk, clipboardy, keysender, sharp, ulid, uuid, zod, @eslint/js
- Some operations may require elevated permissions depending on the target application
load_dotenv() # load environment variables from .env
curl -LsSf https://astral.sh/uv/install.sh | sh
Gates applied: no_behavioural_pass.
ac93f17d1992full audit observations/trust-audit/mcp-server/cheffromspace__mcpcontrol.json · Report an issue / request a re-scanAudit history
Every audit this server has had. A grade with a past is a grade somebody is still checking.
| Date | Source | Verdict | Grade | Score | Change |
|---|---|---|---|---|---|
| 2026-10-07 | ac93f17d1992 | SAFE | B | 89 | first audit |
Questions
What is the MCPControl MCP server?
MCP server for Windows OS automation
What tools does MCPControl expose?
23 in total: 18 read-only, 4 that write, and 1 that can delete or overwrite (clear_clipboard). Every one is listed on this page with its risk.
Is MCPControl safe to connect to an agent?
The audit found nothing in the source that contradicts what it says it does, and graded it B (89/100). Grade A is held back for packages that have also passed a sandboxed behavioural run, which is why a clean server reads B. Separately from the audit: 1 of its tools can destroy data, so scope the token you give it to what you actually need.
What credentials does MCPControl need?
It reads AUTOHOTKEY_PATH and AUTOMATION_KEYBOARD_PROVIDER from the environment. Give it a token scoped to the least it needs — an agent that can be talked into calling a tool can be talked into calling it with your credentials.
How does MCPControl run?
It speaks sse and stdio, so it runs as a local process your client starts. It is published on npm as mcp-control at 0.2.0.
How current is this page?
The grade is for one exact copy of the source (ac93f17d1992), read on 2026-10-07. The repository is watched and re-audited when it changes.