The problem: Most tab managers use simple heuristics—group by domain, or manual tagging. But when I'm researching software options, I'll have the vendor site, pricing articles from different blogs, reviews on Reddit, and demo videos on YouTube—all related but scattered across different domains.
TabSmart uses AI to actually understand what your tabs are about, not just where they're from. It analyzes page content (title + content extraction) and groups tabs that are semantically related, even if they're from completely different sites.
How it works: - Extracts content from open tabs (configurable token limits) - Sends to LLM (OpenAI, Anthropic, etc.) with a prompt to find relationships - Returns structured groups based on semantic similarity - Creates Chrome tab groups automatically
Technical details: - Built as Chrome extension (manifest v3) - Supports BYOK (bring your own API key) or managed service - Customizable AI prompts for grouping logic - Auto-group feature for continuous organization - Privacy-focused: all processing happens server-side (or with your own key)
The interesting challenge was getting AI grouping to be both accurate and fast. Early versions would take forever with 50+ tabs. I optimized by batching tabs, using smarter content extraction, and caching.
You can try it here: https://chromewebstore.google.com/detail/tabsmart-ai-powered...
Built with vanilla JS, Supabase for auth/subscriptions, and Stripe for payments. Happy to answer questions about the implementation!