How Quorly works

Plain HTML, CSS and JavaScript. No framework, no build step, no bundler, no server-side rendering. The same files are served to everyone; the editor is code your browser runs, not a service your browser talks to. That is also why the app works offline after the first load: there is nothing to fetch per keystroke.

Scroll the table sideways to read every column →

There is no account record and no server-side copy on the free plan. Clearing site data deletes the local library — which is the trade-off: nothing can leak from a server that does not hold it, and nothing can be recovered from one either. The privacy page written for writers lists the backup routes.

The only server-side API is /api/*, and it exists for two jobs: knowing which plan an account is on, and deciding which features are enabled. It stores subscription state and flags. It has no endpoint that accepts a title, a paragraph or a file.

• Sign-in for Pro uses Supabase; billing uses Dodo. Both are about money and entitlement, not documents.
• Pro Cloud Sync mints a short-lived Google Drive token. The browser uploads the document directly to your Drive; Quorly brokers the credential and never receives the file.
• Local Sync writes .quorly files to a folder you pick, through the browser's file APIs. No server involved at all.

Exactly four things, and none of them carries document text:

1. Assets from the same origin — the page, the stylesheet, the scripts, the fonts.
2. Cloudflare Web Analytics, which counts visits without cookies or cross-site identifiers.
3. Google Analytics, for aggregate traffic. It may set cookies; the privacy page says so plainly rather than claiming otherwise.
4. A cookieless event to /e — an allowlisted name such as first_keystroke or export, plus at most one short label like the export format. Never text.

Open DevTools → Network and type a paragraph. The request list stays quiet.

The editor ships no model, no assistant and no endpoint a model could be pointed at. Adding one would mean building a service that receives drafts — a change large enough to appear in the network tab on first use. The no-AI page turns that into three questions anyone can ask of any writing app.

In browsers that support WebMCP, the editor does register tools that a local agent in your session can call. That is the opposite direction: an agent you already run can work against the page, the way it could against any tab, and nothing is sent anywhere.

• Longevity. Static files and plain text outlive frameworks. An export today opens in twenty years.
• No lock-in. Markdown, PDF, EPUB and .quorly — the first three are standards, and the fourth is a zip of HTML you can open.
• Verifiability. Claims about privacy are worth what you can check. This design is checkable in a minute.
• Cost. There is no document storage to pay for, because there are no documents on the server.