Bookmarks by Localforge

A knowledge discovery tool for your saved web pages

> Uses your browser to capture a snapshot of the DOM and process it with Mozilla's readability.js

> Semantic search powered by embeddings

> 100% local storage. Optional self-hosted server for multi-device sync.

Download for Firefox Download for Chrome See Source Code
Capture
One-click save with full page snapshot
Search
Find bookmarks by meaning, not keywords
Stumble
Rediscover forgotten bookmarks at random
Organize
Flat tags, bulk import, export anytime

Screenshots

Popup - Quick capture
Popup
Library - Browse bookmarks
Library
Search - Semantic queries
Search

Features

Content Extraction
Converts HTML to clean Markdown using Mozilla's Readability engine
AI-Generated Q&A Pairs
LLM generates 5-10 smart question-answer pairs about each bookmark for RAG search
Tag-Based Organization
Flat organization with tags. Click to filter, type to create new tags instantly.
Stumble Mode
Randomly surface bookmarks you may have forgotten. Filter by tags to stumble within a topic.
System Health Indicator
Visual status in header shows processing state. Click to see diagnostics and job history.
Bulk Import
Import multiple URLs at once with validation and progress tracking.
Export & Backup
Export your full collection or individual entries.
Optional Server Sync
Self-host for multi-device sync with Deno or Cloudflare Workers. Token-based authentication, no passwords.

How It Works

1. Capture — Click the extension or press the shortcut. Full page HTML is saved locally.
2. Extract — Main content is extracted and converted to clean, readable Markdown.
3. Generate — The configured LLM creates Q&A pairs about the content.
4. Embed — Each Q&A pair is converted to vector embeddings for semantic search.
5. Discover — Search by meaning, browse by tags, or stumble upon forgotten gems.

Privacy First

> All bookmarks stored in your browser's IndexedDB — works fully offline.

> Only extracted Markdown is sent to your configured API for processing.

> You control the API endpoint — use local models with vLLM or llama.cpp for full privacy.

> Optional server sync is self-hosted — your data stays on your infrastructure.

> No default queries to any server, and no telemetry.

Read Full Privacy Policy →