Anthropic have introduced the ability to build, host, and share interactive AI-powered apps directly in the Claude app, allowing developers to iterate faster without managing scaling complexity and costs.
The new feature creates artifacts that interact with Claude through an API, transforming them into AI-powered applications with user-friendly economics. When users access shared Claude-powered apps, they authenticate with existing Claude accounts and their API usage counts against their own subscription rather than the developer's account. Developers pay nothing for user consumption and no API key management is required.
Early adopters have built AI-powered games with NPCs that remember conversations and adapt to player choices, learning tools that adjust to individual skill levels for personalized tutoring, data analysis apps where users upload CSVs and ask natural language follow-up questions, writing assistants for scripts and technical documentation, and agent workflows orchestrating multiple Claude calls for complex tasks.
The development process begins by describing desired functionality to Claude, which writes the corresponding code. Claude can debug and improve its own code based on developer feedback. Once completed, apps can be shared instantly through links without deployment processes. Claude handles technical details including prompt engineering, error handling, and orchestration logic.
Current capabilities include using Claude API within artifacts, processing files and creating rich UIs with React, and seeing, forking, and customising any artifact. Current limitations include no external API calls, no persistent storage, and restriction to text-based completion API functionality.
The feature is available in beta to Free, Pro, and Max plan users, enabling rapid AI application development without infrastructure management. Organisations can create custom AI tools for internal workflows while leveraging user-based API consumption models.
The user-pays model eliminates developer infrastructure costs for AI app distribution, potentially accelerating AI application adoption across business segments. The instant sharing capability reduces traditional software deployment barriers while maintaining cost predictability for both developers and end users.