Orellius Eye

Visual perception MCP server — screenshots, UI inspection, element detection.

Website · GitHub

---

What it does

Gives AI agents visual perception of running desktop applications. Takes screenshots, inspects UI trees, finds elements by text or role, and reports window geometry. Works across Electron, Tauri, Flutter, Qt, and SwiftUI apps.

Install

npx orellius-eye

Or globally:

npm install -g orellius-eye

Quick start

Add to your MCP config:

{
  "mcpServers": {
    "eye": {
      "command": "npx",
      "args": ["orellius-eye"]
    }
  }
}

Features

• 6 tools: screenshot, detect_app, inspect_ui, find_element, get_window_info, list_windows
• CDP (Chrome DevTools Protocol) + accessibility API support
• Cross-platform: macOS, Linux, Windows
• Works with Electron, Tauri, Flutter, Qt, SwiftUI apps
• Image optimization via sharp
• Element detection by text, role, or CSS selector

Tech stack

TypeScript, MCP SDK, chrome-remote-interface, sharp

License

MIT — Orellius Labs