Nexu-io launched html-anything on GitHub on May 11, 2026, attracting 857 stars within three days and appearing in GitHub's trending AI projects. The tool represents a new category of development software: HTML editors designed for AI agents rather than human developers, with the tagline "your local AI agent writes the HTML, you ship it."
Supports 75 Composable Skills Across 9 Output Surfaces
The editor integrates with major AI coding assistants including Claude Code, Cursor, Codex, Gemini, Copilot, OpenCode, Qwen, and Aider. It organizes 75 composable skills across nine distinct output surfaces: magazine layouts, presentation decks, posters, social media content for Xiaohongshu and Twitter, prototypes, data reports, and interactive hyperframes. Each surface addresses specific production needs, from creating social media posts to building complete data visualizations.
Zero API Key Architecture Reuses Existing Authentication
Unlike traditional AI tools requiring separate API credentials, html-anything employs a zero API key architecture that reuses authentication from existing coding agents. This "Bring Your Own Keys" (BYOK) approach eliminates setup friction and leverages credentials developers already manage. The tool provides one-click export functionality to WeChat, X (Twitter), Zhihu, HTML, and PNG formats, addressing the practical challenge of moving AI-generated designs into production channels.
Built with Next.js for Local-First Development
The technical stack centers on Next.js with HTML and CSS, emphasizing local-first architecture. The sandboxed preview environment allows developers to test outputs before export. GitHub topics associated with the project include agent-skills, agentic, ai-agents, ai-design, claude-skills, coding-agents, generative-ai, and vibe-coding, positioning it at the intersection of multiple AI development trends.
Key Takeaways
- html-anything gained 857 GitHub stars in three days after launching May 11, 2026
- Supports eight major AI coding assistants including Claude Code, Cursor, and Gemini
- Features 75 composable skills organized across nine output surfaces from magazine layouts to social media posts
- Zero API key architecture reuses existing AI assistant authentication instead of requiring separate credentials
- Provides one-click export to WeChat, Twitter, Zhihu, HTML, and PNG formats for production deployment