Microsoft’s Build 2026 conference delivered a quiet bombshell for Windows developers: the public preview of Windows Development Skills, an AI-driven toolkit that lets coding agents like Anthropic’s Claude Code generate full WinUI 3 applications. Within hours of the announcement, veteran tech journalist Paul Thurrott put it to the test, coaxing Claude into building a working Notepad clone—no hand-coding required.

This isn't another chatbot demo. It’s a glimpse of a future where describing an app in plain English produces a packaged, natively compiled Windows program in minutes. The magic comes from a new command-line tool called WinApp CLI, a scaffolding engine that arms AI agents with the precise steps needed to create, configure, and deploy Windows App SDK projects.

What is WinApp CLI?

WinApp CLI is Microsoft’s answer to the friction between AI coding assistants and the complex reality of Windows development. Traditional AI tools can generate code snippets, but they often stumble on project setup, package manifests, and platform-specific plumbing. WinApp CLI solves this by encapsulating every Windows App SDK workflow—project creation, dependency management, packaging, signing, and deployment—into a small set of deterministic commands.

Under the hood, WinApp CLI speaks the language of dotnet new templates, MSIX packaging, and the Windows App Runtime. But instead of developers having to remember these arcane incantations, the AI agent calls commands like winapp new to scaffold a WinUI 3 project, winapp package to bundle it as an MSIX installer, and winapp run to launch it on the local machine. Each command produces predictable outputs that an LLM can parse and learn from.

Crucially, WinApp CLI is not just a launcher. It maintains a project’s entire lifecycle. Need to add a feature? The AI can edit the XAML and C# files, then call winapp build to verify the result. Want to test on a different architecture? winapp build --arch x64|arm64 handles that. This deterministic bridge is what turned Claude Code from a generalist coder into a competent Windows developer.

Windows Development Skills: Teaching AI the Rules of the Road

Even with perfect tooling, an AI needs context. That’s where Windows Development Skills come in. Announced as a preview alongside the new WinApp CLI, Windows Development Skills is a curated collection of knowledge, best practices, and code patterns that ground AI models in the specifics of the Windows platform.

Think of it as a system prompt on steroids. When an AI coding agent is equipped with Windows Development Skills, it gains an understanding of WinUI 3’s control library, the MVVM pattern, the navigation model, data binding, and even accessibility guidelines. It knows when to use a ListView versus a GridView, how to handle windowing, and why Package.appxmanifest matters.

Microsoft engineered these skills to be model-agnostic. During Build, demos showed both Claude Code and GitHub Copilot taking advantage of them. The skills are delivered as part of the AI agent’s context window, meaning the model doesn’t need to hallucinate the Windows API—it can ground its responses in verified schemas and samples.

For Thurrott’s experiment, he combined the preview of Windows Development Skills with the Anthropic Claude Code CLI. The result was an agent that could not only generate a WinUI 3 Notepad app but also understand the structural conventions of a modern Windows application.

From Prompt to Working Notepad Clone in Minutes

Thurrott’s experiment began with a deliberately vague prompt: “Create a Notepad app for Windows 11.” He then stepped back as Claude Code invoked WinApp CLI to scaffold the project, studied the Windows Development Skills context, and started generating code.

What emerged was a fully functional text editor. It supported standard editing operations—cut, copy, paste, undo, redo—displayed a title bar with the file name, and even included a menu bar with File, Edit, and Help menus. The window behaved like any native Windows 11 application: it snapped into snap layouts, integrated with the taskbar, and respected the system theme (light or dark).

“It wasn’t just throwaway demo code,” Thurrott reported. “The XAML was clean, the view model was properly separated, and it compiled without a single error.” The entire process, from prompt to running application, took under three minutes on a modest development machine.

Of course, a Notepad clone is not a production app. But the speed and accuracy hint at something profound. The AI didn’t just regurgitate a template; it made sensible design choices. For example, it chose a RichEditBox control over a plain TextBox to handle larger files, and it implemented basic file I/O using the Windows Storage APIs—exactly the approach a seasoned Windows developer would take.

Why This Matters for Windows Development

The ability to generate a native Windows application from a natural language description addresses long-standing pain points. First, it drastically lowers the barrier to entry. WinUI 3 is powerful, but its learning curve is steep. By offloading boilerplate and configuration to AI, newcomers can focus on their app’s unique logic rather than wrestling with XAML layout and package manifests.

Second, it accelerates prototyping. Product managers, designers, and even business analysts can generate functional UI mockups that developers can then refine. That shortens the feedback loop from clickable prototype to real code to hours instead of weeks.

Third, it modernizes legacy Windows development. Many enterprises maintain sprawling WinForms or WPF codebases. Windows Development Skills could help teams incrementally migrate to WinUI 3 by generating reference implementations and mapping legacy patterns to modern equivalents.

However, the technology is not a silver bullet. Complex apps with state management, custom controls, or hardware integration still require human judgment. The generated Notepad app, for instance, lacked features like syntax highlighting, find-and-replace, or printing—real-world capabilities that demand deep domain knowledge. And while the code was syntactically correct, a thorough code review would still be necessary to catch subtle security or performance issues.

Community Reaction: Excitement Tempered with Skepticism

Within hours of the Build session, Windows development forums lit up. Enthusiasts praised the demos but raised familiar concerns. “Will this actually produce maintainable code, or just a fresh kind of spaghetti?” one developer asked. Others worried that the technology might encourage copy-paste culture, leading to a flood of low-quality apps in the Microsoft Store.

Microsoft has been careful to position Windows Development Skills as an aid, not a replacement. “We’re not trying to automate developers out of a job,” a program manager said during the Build Q&A. “We’re automating the parts that should have been automated a decade ago—scaffolding, packaging, signing. The creative work remains human.”

Early adopters echoed that sentiment. “It’s like having a junior dev who is incredibly fast but needs supervision,” noted one community member who tried the preview. “You still need to understand what good code looks like. But for tedious setup, it’s a game changer.”

The Broader AI Coding Landscape

Microsoft’s move fits into a broader industry trend where every major platform is rushing to integrate AI coding agents. GitHub Copilot Workspace, Google’s Project IDX, and various open-source tools all aim to turn prompts into software. What sets the Windows effort apart is its deep platform integration.

By giving AI agents first-class access to Windows App SDK internals through WinApp CLI, Microsoft ensures the generated code aligns with official guidance on security, performance, and accessibility. That’s a stark contrast to generic coding assistants that might produce code that runs but violates Windows certification requirements.

Additionally, Windows Development Skills can be extended. Enterprise teams can author custom skills that teach the AI about their proprietary frameworks, design systems, or internal APIs. That means a company could train its agent to generate screens that adhere to its branding guidelines or integrate with its back-end services out of the box.

What’s Next for WinApp CLI and AI on Windows

Microsoft has revealed that WinApp CLI will graduate from preview to general availability later this year, with support for additional project types beyond WinUI 3, including desktop widgets and Win32 interop scenarios. The company is also working on tighter integration with Visual Studio and VS Code, allowing human developers to seamlessly switch between AI-generated code and manual edits.

Perhaps most intriguing is the potential for on-device AI. As Windows devices increasingly ship with neural processing units (NPUs), future iterations of Windows Development Skills could run locally, preserving privacy and eliminating cloud latency. Imagine dictating an app idea to your laptop and watching it take shape in real time—all without sending a byte of code to the internet.

For now, the immediate takeaway is clear: building a native Windows application no longer requires a deep understanding of XAML, C#, or the Windows App SDK. It requires a good idea and the ability to describe it clearly. The AI handles the rest.