OpenSky
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@OpenSkytake a screenshot of the active window and describe it"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
OpenSky
Universal Computer Use Engine & MCP Server
English | 简体中文
OpenSky is a cross-platform Computer Use engine and Model Context Protocol (MCP) server designed to enable AI agents (such as OpenCode, Reasonix, Claude Desktop, Cursor, Antigravity, and OpenAI Codex) to automate desktop operating systems and applications.
The project currently focuses on providing a native Windows driver, while maintaining a platform-agnostic interface layer to allow future drivers for macOS and Linux.
Key Design Principles
1. Dual-Track Interaction
UI Automation (Accessibility Tree): Parses the application control tree and assigns sequential numeric indices (e.g.,
[15] Button "Submit"). Agents can interact directly by index (win_click(elementIndex: 15)orwin_set_value(elementIndex: 15, value: "...")). This approach is independent of display scaling (DPI) and eliminates the need to estimate pixel coordinates from images.Visual Perception (Screenshots & Coordinates): Provides full-screen or window-specific screenshot capabilities. Agents can fall back to coordinate-based clicks and drags when interacting with non-standard controls, canvases, or web views.
2. Visual Feedback Overlay
Includes an overlay window built using Win32 layered transparent windows (WS_EX_TRANSPARENT | WS_EX_LAYERED):
Click Ripples: Displays a temporary expanding ring animation at the click position to indicate interaction points.
Window Highlight: Outlines the active target window.
Input Pass-Through: The overlay does not capture focus and does not interfere with user mouse and keyboard input.
3. Decoupled Architecture
Core Engine (
src/core): Defines the platform-independentIComputerUseEngineinterface and data models.Transports (
src/driver/transport): Inter-process communication abstraction (STDIO, Named Pipe).Platform Drivers (
src/driver):Windows: Split into modular micro-services for window enumeration, screen capture, accessibility tree parsing, input synthesis, and overlay rendering. Compiled using the system built-in
csc.exe.macOS / Linux: Interface ready for future platform driver implementations.
Adapters (
src/adapters): Includes the standard MCP server, Anthropic Computer Use bridge, and Codex plugin exporter.
Related MCP server: Windows MCP Server
Architecture
+-------------------------------------------------------------+
| AI Agent / Client |
| (OpenCode, Reasonix, Claude Desktop, Cursor, Codex, etc.) |
+------------------------------+------------------------------+
|
+------------------------------v------------------------------+
| Adapters Layer |
| - MCP Server (JSON-RPC) |
| - Anthropic Bridge (computer_20241022) |
| - Codex Plugin Exporter |
+------------------------------+------------------------------+
|
+------------------------------v------------------------------+
| Core Engine Interface |
| (IComputerUseEngine) |
+------------------------------+------------------------------+
|
+---------------+---------------+
| |
+--------------v---------------+ +--------------v---------------+
| Windows Driver | | macOS / Linux (Planned) |
| (WinComputerUseHelper) | | (Accessibility / AT-SPI) |
| - Window Management | +------------------------------+
| - Screen Capture |
| - UI Automation Tree |
| - Input Simulation |
| - Feedback Overlay |
+------------------------------+Requirements & Building
Requirements
Windows 10 / 11 (64-bit)
Node.js >= 20.0.0
pnpm
Built-in .NET Framework 4.8 (
csc.exelocated inC:\Windows\Microsoft.NET\Framework64\v4.0.30319\)
Build Steps
# Install dependencies
pnpm install
# Compile C# native helper and bundle TypeScript code
pnpm run build
# Run automated tests
pnpm testClient Configuration
OpenCode
Add OpenSky to opencode.jsonc:
{
"mcp": {
"opensky": {
"type": "local",
"command": ["node", "/path/to/opensky/dist/bin/opensky.js"]
}
}
}Reasonix / Google Antigravity / Cursor
The repository includes a .mcp.json file in the root directory, which is automatically recognized by supported clients:
{
"mcpServers": {
"opensky": {
"command": "node",
"args": ["./dist/bin/opensky.js"]
}
}
}Claude Desktop
Add to %APPDATA%\Claude\claude_desktop_config.json:
{
"mcpServers": {
"opensky": {
"command": "node",
"args": ["/path/to/opensky/dist/bin/opensky.js"]
}
}
}OpenAI Codex
Export OpenSky as a Codex plugin directory:
node dist/bin/opensky.js --export-codex ./opensky-codex-pluginThis generates .codex-plugin/plugin.json, .mcp.json, and skills/opensky/SKILL.md in the target directory.
Tool Reference
Tool Name | Parameters | Description |
| None | Lists visible application windows with HWND, title, process name, and rectangle. |
|
| Brings target window to foreground and outlines it. |
|
| Captures full desktop or target window, returning Base64 JPEG. |
|
| Dumps accessibility tree with sequential element indices and discovered supported patterns. |
|
| Zero-Cursor: Directly triggers UI Automation Pattern (Invoke, Toggle, Select, SetValue) in memory without moving physical mouse or stealing focus. |
|
| Clicks an indexed element or screen coordinate, triggering a visual ripple. |
|
| Performs a double click. |
|
| Performs a right click. |
|
| Performs smooth mouse drag from start to end coordinates without screenshots. |
|
| Types Unicode text into the currently focused control. |
|
| Sends key or hotkey combination (e.g. |
|
| Scrolls the mouse wheel at the specified position without screenshot overhead. |
|
| Sets control text directly using UI Automation ValuePattern. |
License
This project is licensed under the MIT License.
This server cannot be deployed
Maintenance
Related MCP Connectors
- mcp-serverOAuthcom.make
Give your AI agents the tools to build, manage, and run automation workflows.
Control real Android and iOS devices with LLM agents — tap, swipe, type, automate flows.
Operate Linux, macOS and Windows from your LLM. Every action runs through an auditable allowlist.
AI-powered web automation. Navigate websites using AI agents for one page or a thousand
Related MCP Servers
- AlicenseNot gradedqualityCmaintenanceEnables AI agents to interact with Windows operating systems through native UI automation, file navigation, application control, and system commands. Provides seamless integration between LLMs and Windows environments for tasks like clicking, typing, launching apps, and capturing desktop state.MIT
- AlicenseNot gradedqualityBmaintenanceEnables comprehensive Windows desktop automation including screen capture, OCR text extraction, mouse/keyboard control, window management, process control, and clipboard operations through 25+ tools for AI agents.46 PyPI5MIT
- FlicenseNot gradedqualityCmaintenanceEnables AI agents to control Windows GUI applications like a human using screen capture, OCR, mouse and keyboard input, and window management, with safety levels and memory.-
- AlicenseNot gradedqualityDmaintenanceEnables AI agents to interact with the Windows operating system, performing tasks such as file navigation, application control, UI interaction, and QA testing.MIT