You have 47 tabs open.
You'll never read most of them.

TabReset summarizes your tabs to a Markdown file, then closes them. Uses local LLMs - no data leaves your machine.

The tab problem

Tabs become a to-do list you never asked for. Each one is something you meant to read or come back to. Closing them feels like giving up. So they pile up until your browser slows down and finding anything becomes impossible.

TabReset gives you permission to close tabs by making sure you won't lose them. One click: summarized, saved, closed.

How it works

1
Pick a file
Choose where to save your tab archive. It's a Markdown file. Open it in any text editor, Obsidian, or Notion.
2
Click "Summarize & Reset"
TabReset reads each tab, generates a one-line summary, and appends it to your file with the title, link, and timestamp.
3
Tabs are logged and closed
Your browser is clean. Your archive is searchable. When you need that article from three weeks ago, you can find it.

What your archive looks like:

          
                Title
                Summary
                Link
                Closed
              
                React Server Components
                Explains RSC architecture and when to use them
                [link]
                2026-01-21 14:30
              
                Why we moved to Postgres
                Team's migration from MongoDB, lessons learned
                [link]
                2026-01-21 14:30
              
                CSS Container Queries
                How container queries differ from media queries
                [link]
                2026-01-21 14:30

Title	Summary	Link	Closed
React Server Components	Explains RSC architecture and when to use them	[link]	2026-01-21 14:30
Why we moved to Postgres	Team's migration from MongoDB, lessons learned	[link]	2026-01-21 14:30
CSS Container Queries	How container queries differ from media queries	[link]	2026-01-21 14:30

Your tabs stay on your machine

The AI runs locally in your browser using a small language model (LFM2-350M). No servers, no accounts. Nothing leaves your computer.

The model downloads once (about 700MB), then everything happens offline. Your tab contents are never transmitted anywhere.

Local processing

Summarization happens in-browser via WebAssembly. The model runs on your CPU.

Your files, your location

You pick where to save. Desktop, Dropbox, a git repo. Wherever works.

Append-only

TabReset only adds to your file. It never reads, deletes, or modifies existing content.

Under the hood

TabReset prioritizes speed. Before running the LLM, it checks for meta descriptions and Open Graph tags. If a page already has a good summary, it uses that instead. For pages without metadata, it extracts article content and runs it through the model.

Built on Transformers.js and the Chrome File System Access API. The extension is open source.

Close your tabs. Keep the knowledge.