Ever wish you could whip up a little AI sidekick or a custom flashcard app without wrestling with boilerplate code and API keys? That’s exactly the promise Anthropic just doubled down on. On June 25, the company rolled out an interactive beta in its Claude chatbot that turns your casual prompts into fully fledged, shareable apps—no external IDE required and no surprise cloud bills gnawing at your wallet.

Last year’s Artifacts feature let you “see and interact” with whatever Claude whipped up—be it a to-do list generator or a snippet of JavaScript. Now, those static results can morph into living applications. All you do is toggle on “Create AI-powered artifacts” in your Claude settings, describe what you want, and watch Claude draft real code right in the chat window. A quick demo video from Anthropic shows someone asking for a chat app, then immediately chatting with their creation—all without leaving Claude.

At its core, this update introduces a deep integration between Claude’s conversation engine and a lightweight in-app runtime sandbox. When you tell Claude to “build a flashcard app,” it generates HTML, CSS, and JavaScript wrapped in an <iframe> with strict content-security policies. Behind the scenes, your app can invoke window.claude.complete(), firing off additional Claude calls via a new embedded API—so your flashcards can dynamically fetch definitions or quiz questions.

No more hunting for API keys or piecing together SDKs. Each end user simply authenticates with their own Claude account, and any AI usage they trigger ticks against their subscription—not yours. Anthropic’s blog promises zero hosting costs for creators, since “you pay nothing for their usage, and no one needs to manage API keys.”

Behind the scenes, these creations live in a new “Artifacts” tab inside Claude, complete with browsing, forking, and customization tools. Anthropic reports that its half-billion existing artifacts—from quick tip generators to code snippets—now have a shot at interactive life.

Anthropic isn’t just handing out wrenches; it’s building the garage. Shared artifacts live on Anthropic’s platform, and when your friend—or tomorrow’s user—runs your app, the compute they consume is billed to them. This model removes a major barrier for indie creators, students, and small teams who want to experiment without surprise invoices or complex cloud setups.

The company’s encouragement to “share your creations” highlights a shift toward community-driven innovation. Think of it as GitHub for AI widgets, where “pull requests” come in the form of “Claude forks” and issue trackers are just follow-up chat threads.

Anthropic’s move follows similar pushes from the likes of OpenAI (with its ChatGPT Plugins) and Microsoft’s Copilot Canvas, but with a unique twist: everything happens within a single chatbot interface. No separate dev console, no SDK installation. It’s a pure “vibe coding” experience—describe your vision, and your code appears.

Some veteran developers might bristle at the lack of manual control or debugging hooks. Yet for many creators—educators, product managers, hobbyists—the ability to prototype and share AI tools in minutes could unlock entirely new use cases. After all, the easier you make it to build, the more people will explore and iterate.

For now, the feature is in beta and available to all Claude tiers—Free, Pro, and Max—so you can dive in regardless of plan. Anthropic’s roadmap hints at more robust publishing tools, better collaboration workflows, and deeper customization options. In time, these in-chat apps could spawn independent micro-SaaS projects or classroom tools that teachers build and distribute with no coding background.