openeyes

v0.1.0

OpenEyes — AI-friendly computer-use primitives (capture/detect/click) as a CLI and MCP server. Drives any desktop GUI on Windows/macOS/Linux via the platform accessibility tree, with Vimium-style grid fallback.

Codex1 Skill

By yangpengLicense: MIT0 GitHub starsUpdated 2 hours ago

Directory evidence

Runtimes: Codex
Parsed components: 1 skill or MCP entry
Source updated: Aug 29, 2026
Manifest status: Canonical path parsed

The directory validates manifest shape and source location. It does not execute the plugin or provide a security endorsement. Review the indexing methodology →

Install plugin

Installs for the current user

codex plugin marketplace add IchenDEV/agent-plugin-mkt
codex plugin marketplace upgrade agent-plugin-marketplace
codex plugin add openeyes@agent-plugin-marketplace

Paste and run these commands in a terminal with Codex. They add and refresh the PluginsMP catalog, then install this plugin.

The installer fetches third-party code from the source repository shown on this page. This directory validates manifest structure and source location, but does not perform a security audit; review the manifest, components, and source before installing.

Get the source manually

$ git clone https://github.com/yangpeng366/openeyes

Clone the source repository, then follow its setup instructions to add the plugin to a compatible client. The repository root is the plugin root.

Plugin files

openeyes/

├── .codex-plugin/plugin.json
└── skills/openeyes/SKILL.md

Included Skills1

openeyesskills/openeyes/SKILL.md

Use OpenEyes (eyes CLI / openeyes Python package / openeyes-mcp server) to see and operate any desktop UI on Windows/macOS/Linux. Includes UIA-backed native apps (Win32/UWP), CDP-backed Chromium browser pages (Edge/Chrome via DevTools Protocol — no screenshot, no vision model, structured DOM), Vimium-style letter hints, Vimium-style 3x3 grid, focus/click/type/hotkey/drag/screenshot. Use when Codex needs to (a) drive a native app without an API or DOM, (b) drive a real browser session with login cookies / 2FA, or (c) read or write page elements (text input, button click) without spending image-budget. Browser path uses raw websocket-client on Chrome DevTools Protocol; Vision backends (OmniParser / Florence-2) remain as Phase-3 fallback for canvas / image-only UIs.

Plugin manifests1

.codex-plugin/plugin.json

{
  "name": "openeyes",
  "version": "0.1.0",
  "description": "OpenEyes — AI-friendly computer-use primitives (capture/detect/click) as a CLI and MCP server. Drives any desktop GUI on Windows/macOS/Linux via the platform accessibility tree, with Vimium-style grid fallback.",
  "author": {
    "name": "yangpeng",
    "email": "[email protected]"
  },
  "homepage": "https://github.com/yangpeng366/openeyes",
  "license": "MIT",
  "keywords": [
    "openeyes",
    "computer-use",
    "ai-agent",
    "uia",
    "screenshot",
    "click",
    "mcp",
    "vision",
    "automation"
  ],
  "skills": "./skills/openeyes/",
  "interface": {
    "displayName": "OpenEyes (Computer-Use)",
    "shortDescription": "Capture/detect/click + MCP server for AI desktop automation",
    "longDescription": "OpenEyes provides primitives for AI agents to see and operate any desktop UI: capture screenshots, enumerate interactive elements via the platform accessibility tree (UIA on Windows, AX on macOS, AT-SPI on Linux), click by coordinates / text / handle / hint, snap to a Vimium-style grid, drag, type, send hotkeys. Pluggable vision backend (OmniParser/Florence) for legacy UIs. Safe-by-default: every click defaults to dry-run.",
    "developerName": "yangpeng",
    "category": "Developer",
    "capabilities": [
      "Interactive"
    ],
    "defaultPrompt": [
      "Use openeyes to find the message input box in the Feishu client and click it",
      "Take a screenshot of the foreground window, enumerate its clickable elements, and click the one matching the user's intent",
      "Drive an unknown legacy desktop app by enumerating its controls via UIAutomation",
      "Open the Any VPN window and re-click the 'Secure my connection' button every 20s (dry-run by default)"
    ],
    "brandColor": "#0EA5E9",
    "screenshots": []
  }
}

For maintainers

If you maintain this plugin, link to this source-backed listing from your README so users can review its manifest and indexed components.

[openeyes on Agent Plugins Marketplace](https://pluginsmp.com/plugins/openeyes)