Read-later services keep shutting down and taking people's libraries with them. A local-first version you own solves that, and it's a good exercise in search and data portability.

1. Save more than the URL

The title, the URL, tags, a note, when you saved it, and ideally a short excerpt. That last one is what makes search actually work — searching only titles misses most of what you'd look for.

bookmark-manager.txt
Build a local-first bookmark manager. Browser storage, no account.

- Add by URL, fetching the title and description via <a small proxy /
  manual entry — recommend one and say why fetching cross-origin from
  the browser won't work>.
- Tags, with autocomplete from existing tags.
- Full-text search across title, URL, tags and notes. Instant, as I
  type.
- Filter by tag, sort by date or title.
- Import from the browser's own bookmarks HTML export, and export back
  to the same format.

Handle a few thousand bookmarks without the search getting slow.

2. Fetching titles needs help

The browser can't fetch an arbitrary page cross-origin and read it — that's what CORS prevents. Your options are a small server-side endpoint that fetches and extracts metadata, a browser extension, or asking the user to type the title.

Be honest about this in the design rather than discovering it halfway through.

3. Search is the product

A bookmark you can't find again is a bookmark you didn't save. Build the search index up front rather than filtering an array — at a few thousand items, filtering on every keystroke starts to lag.

4. Import and export as first-class

Import from the browser's bookmark export so people can start with what they have. Export in the same format so they can leave. That's what makes it trustworthy, and it's also your backup path.

Store the date you last opened each bookmark. A "saved 400 things, opened 12" view is uncomfortable and genuinely useful.