htmlbook - A bookshelf for the HTML & MD files your AI agents generate
by•
Your AI writes HTML — htmlbook keeps it. Capture documents from Claude Code, Cursor, Codex over MCP, organize them into a themed library, and share any page with a link. Read anywhere.
Replies
Best
Maker
📌
Hey hunters 👋
I'm nullarch, a solo founder. I basically live inside Claude Code and Cursor, and they generate a constant stream of HTML and Markdown — PRDs, dashboards, research notes, todo lists. They all land as local files, and that one fact created three pains that kept adding up for me:
Sharing is painful. Sending someone a raw .html renders like garbage, and standing up hosting for a doc I'd use once is absurd overkill.
I can't read my own docs anywhere but my machine. No glancing at a PRD on my phone on the train, no pulling up a dashboard on the couch.
Nothing stays organized. Files just pile up across random folders — no structure, no history, no way to find last week's version.
htmlbook fixes all three. It's a bookshelf your agent pushes to. You connect once — one line of MCP in the terminal, or OAuth from claude.ai/Desktop — then just tell the agent "put this on my bookshelf."
And each pain maps to something:
Sharing → flip a doc public and get a short link (htmlbook.io/d/…) that opens on any phone, no install or login for the person receiving it.
Reading anywhere → plain HTML/Markdown gets a reader with paper/sepia/dark themes + real typography, so it's actually nice on a phone. Pages with their own CSS/JS render as-is, sandboxed.
Organization → docs auto-sort into projects / workspaces / tags, and every version is kept immutably.
There's deliberately no web editor — the web side is read/organize only, and the agent push is the main path. That constraint is the whole point.
Free to start: signup gives you a bookshelf + API key, ~1 min to connect.
Genuinely curious — how are you handling the HTML/MD your coding agents produce right now? Screenshotting into Slack? Gists? Or letting it pile up locally like I was?
Replies