MCP Server Security Model
Trust boundaries, permissions and tool-call risks.
A knowledge tool that keeps your files yours
HTMlore gathers the HTML files spread across your disk into one place — a reader and a knowledge base at the same time. AI keeps extending, iterating and settling what you learn, and a free static host turns any note into a shareable page that builds your personal brand.
Reader + Knowledge base
HTML files pile up everywhere — exported articles, AI answers saved to disk, old reports, single-page apps. HTMlore imports them into one library, reads them in a clean card workspace, and organizes them with collections, tags, favorites and search. It is a reader and a knowledge base in the same window.
Files stay portable. Notes remain real HTML on disk — inspect, copy, back up, or serve them with ordinary web tools. Nothing locked in a database.
Organize without friction. Collections, multi-select tags with AND/OR, favorites, archive, sort and full-text search keep a growing library navigable.
Read, don't dig. Open any note full-screen, view the original file, copy or share its link — the reading view is the workspace.
Trust boundaries, permissions and tool-call risks.
Grids, rhythm and hierarchy for dense reading.
One service, persistent data, simple backup.
A living page, extended after every session.
The learning flywheel
A note isn't a dead file. Ask AI over your own library and the answer becomes a new HTML note. Revisit it, refine it, and the page deepens. Each turn feeds the next — extend, iterate, settle — until scattered reading becomes a personal body of knowledge that quietly compounds over time.
Grounded in your library. The AI layer works over notes you choose as context — answers stay tied to what you've actually collected.
Output is a note, not a chat bubble. Generated HTML lands back in the library, ready to read, tag, and grow.
Credentials stay server-side. Model keys live in the backend, never baked into the static frontend.
Save an HTML note — an article, an idea, an AI answer — into your library.
Ask AI across your library; the reply comes back as a new note.
Revisit and refine it; the page deepens with every pass.
It settles into your library — tagged, searchable, and reusable.
Publish & personal brand
Any note can become a public page. HTMlore builds a static site from your content — host it for free on Pages-style infrastructure, send a link, and your work is online. No build pipeline, no server bill. Over time your published notes become a body of work — the quiet foundation of a personal brand.
Static by design. Build a read-only site from existing content and metadata, then serve it anywhere static files live.
Free to host, instant to share. No subscription to publish — copy a link and your page is live for anyone.
Installable as a PWA. A manifest and service worker make the workspace feel like an app on any device.
Polished, fast, mobile-ready HTML — not a database export.
Copy and send. The recipient just opens a URL.
Static output runs on free Pages-style hosts.
A growing set of published notes becomes your portfolio.
In the box
Upload HTML into generated or imported libraries, raw content kept outside Git.
Collections, tags, favorites, archive, search, sort, and AND/OR tag filters.
Full HTML reading, original-file access, copy/share, metadata editing, archive & restore.
Contextual Q&A and note generation grounded in manually selected files.
Built-in login, HttpOnly sessions, per-user notebooks, documented security boundary.
One Docker service, a static build mode, and a PWA — pick what fits.
Two minutes to running
Run the full notebook in a single Docker container on a laptop, NAS, LAN box or private VPS. Or build a static site for free public hosting. The files never leave your control.
git clone https://github.com/JMoCoder/html_lore.git
cd html_lore
docker compose up -d --build
# open http://localhost:8080