Desktop Controller Skill

\

Desktop Controller — AI Computer Use for Claude Code

**The open-source alternative to OpenAI Codex's computer use.** Control any Windows app — native desktop, web, and Electron — with a single AI skill. Like [OpenAI's playwright-interactive](https://github.com/openai/skills/tree/main/skills/.curated/playwright-interactive), but with **native Win32 desktop app support** that Codex can't do.

[](https://opensource.org/licenses/MIT)

**What GPT-5.4 does in the cloud, this skill does locally.** OpenAI's GPT-5.4 introduced "computer use" — the ability to control computers via Playwright and mouse/keyboard commands. Their `playwright-interactive` Codex skill enables visual debugging of web and Electron apps. **We go further**: our dual-engine approach adds **native Windows desktop automation** (Win32 API) alongside Playwright, letting you control apps like WeChat, DingTalk, and QQ that browser-based solutions simply can't reach.

Why This Exists

	OpenAI Codex	Claude Code + This Skill
Web/Electron apps	Playwright	Playwright
Native desktop apps	Not possible	Win32 API
Chat apps (WeChat, DingTalk, QQ)	Not possible	Full support
Visual feedback	js_repl screenshots	Screenshots + Claude Vision
QA workflow	Manual checklist	Automated QA checklist
Viewport testing	Manual	Device presets (iPhone, iPad, Pixel...)
DOM inspection	Via DevTools	Built-in inspector
Platform	Cloud only	Local (your machine)
Cost	Codex subscription	Free & open source

Demo

$ python scripts/desktop_control.py list-apps
Supported applications:
  weixin       WeChat                process=Weixin       mode=win32   search=Ctrl+F
  wxwork       WeCom (企业微信)        process=WXWork