nexu-io/html-anything
✨ The agentic HTML editor — your local AI agent writes the HTML, you ship it. 🚀 75 Skills × 9 Surfaces (magazine · deck · poster · XHS / tweet · prototype · data report · Hyperframes) 🛡️ Sandboxed preview · 📤 1-click to WeChat / X / Zhihu / HTML / PNG 🔑 Zero API key — Claude Code / Cursor / Codex / Gemini / Copilot / OpenCode / Qwen / Aider.
HTML Anything – Agent‑powered HTML editor
What it is – A local‑first desktop‑style web app that lets you write content in any plain‑text format (Markdown, CSV, JSON, SQL, etc.) and, with the help of a locally‑installed AI coding‑agent (Claude, Cursor, Codex, Gemini, Copilot, OpenCode, Qwen, Aider or IBM Bob), instantly turns that input into a complete, single‑file HTML page. The generated page is styled with Tailwind‑based design‑systems and can be exported directly to Chinese social platforms (WeChat, X, Zhihu) or saved as .html/.png.
Why it matters – Modern AI assistants are great at generating code, but most developers still have to copy‑paste the output into a Markdown editor and then convert it to HTML for publishing. HTML Anything skips that middle step: the AI writes the final HTML that readers actually see, and it does so on‑device without needing any API keys. This makes the workflow fast, private, and cost‑free (apart from the user’s existing AI subscription).
Key features
- Agent detection – On start‑up the app scans your
PATHfor nine supported coding‑agent CLIs and lets you pick which one to use. - 75 ready‑made skill templates – Pre‑defined layouts for 9 surface types (magazine article, keynote deck, résumé, poster, Xiaohongshu card, tweet card, web prototype, data report, Hyper‑frames video) and dozens of design variants (Swiss International, Guizang Editorial, etc.).
- Live streaming render – The agent streams JSON‑line output; the front‑end parses it and updates an iframe preview in real time, so you can watch the page being built line‑by‑line.
- One‑click export – Inline CSS (“juice”) for WeChat, high‑DPI PNG screenshots for X/Weibo/Xiaohongshu, Zhihu LaTeX placeholders, or plain
.html/.pngdownload. - Sandboxed preview – The generated HTML runs inside a sandboxed iframe, keeping any scripts isolated from the host.
- Zero API‑key requirement – Re‑uses the authentication session you already have with the chosen agent (e.g.,
claude login). - Deployable locally or on Vercel – Run with
pnpm -F @html-anything/next devor deploy the front‑end; the AI component always stays on the user’s machine.
Typical workflow
- Install one or more supported coding‑agent CLIs and log in.
- Clone the repo, run
pnpm installand start the dev server. - In the browser, pick a skill/template (e.g., a magazine‑style article).
- Paste your source content (Markdown, CSV, etc.) and press ⌘+Enter.
- The local agent generates HTML; you watch it stream into the preview.
- Export with a single click to the desired platform or download the file.
Who should use it – Content creators, marketers, product writers, and designers who want AI‑generated, publication‑ready HTML without manual CSS work, and who already have access to a supported coding‑agent CLI.
Repository layout
next/– The Next.js front‑end (@html-anything/next).e2e/– Playwright end‑to‑end tests (@html-anything/e2e).scripts/guard.ts– CI helper scripts.next/src/lib/agents/– Adapters that translate each CLI’s JSON‑line protocol into a common interface.next/src/lib/templates/skills/– The 75 skill folders, each with aSKILL.mddescriptor and an example HTML file.
License – Apache‑2.0.
All details above are taken directly from the project’s README.
Related
- Project
- Project
- Project
- Project
- Project