Chrome Extension · Gemini Nano · On-device AI

Your tabs,
intelligently managed

Tab Agent groups your open tabs by topic, learns which ones you actually return to, and lets you sleep, wake, or close entire groups with one click. No API key. No data leaves your device.

Install from GitHub See how it works

The problem with tabs

Chrome's built-in tab suspension uses a fixed inactivity timer — it will suspend a Figma tab you return to every 20 minutes just as readily as a news article you opened 3 hours ago and forgot about. When a tab is suspended, Chrome discards the page from memory. Revisiting it means a full reload — network requests, JS execution, re-rendering. For heavy pages like Google Docs or Figma, that's genuinely disruptive.

Existing extensions like The Marvellous Suspender, OneTab, and Tab Wrangler are either rule-based or require manual organization. None of them learn from you.

What Tab Agent does

◈

AI grouping

Gemini Nano groups your open tabs by topic — Work, Research, Shopping — instantly, on your device.

◉

Frequent tab protection

Tabs you visit often get a badge and are never suspended without your confirmation.

◐

Sleep & Wake

Sleep an entire group to free memory. The group stays visible — wake it with one click when you need it back.

◑

Persistent groups

Close and reopen the popup anytime. Groups are remembered — no re-grouping, instant load.

◒

Stats dashboard

See memory saved, tab counts, and awake/asleep status per group. Rate grouping quality for research.

◓

100% private

No API key. No server. Gemini Nano runs entirely in your browser. Your tabs never leave your device.

How the agent works

Observe

Reads all open tabs — title, URL, and your visit history from local storage.

Decide

Sends the tab list to Gemini Nano. Gets back named groups. Caches the result.

Act

You click Sleep, Wake, or Close on a group. The agent never acts without you.

Remember

Logs every tab switch. Tabs visited 3+ times in 24h are marked frequent and protected.

Coming next: The agentic version will sleep and wake tabs autonomously — sleeping low-need tabs based on your behavioral patterns, and waking related tabs when you switch context. The user sets the policy once; the agent handles the rest.

How to install

Tab Agent isn't on the Chrome Web Store yet — install it in developer mode in 3 steps.

Enable Gemini Nano

Open Chrome and go to these two URLs. Set the values shown, then click Relaunch.

chrome://flags/#prompt-api-for-gemini-nano
→ Set to: Enabled

chrome://flags/#optimization-guide-on-device-model
→ Set to: Enabled BypassPerfRequirement

Then open DevTools on any page and run:

await LanguageModel.create()

This downloads the model (~3GB). Wait for it to finish, then verify with await LanguageModel.availability() — should return "available".

Download the extension

Clone or download the repo from github.com/MaykaS/tab_agent

git clone https://github.com/MaykaS/tab_agent.git

Load in Chrome

Go to chrome://extensions, enable Developer mode, click Load unpacked, and select the tab_agent folder.

Tab Agent appears in your toolbar. Click it to group your tabs.

Research project

Tab Agent is a graduate research project at Cornell Johnson MBA, exploring agentic systems for browser memory management. The project studies three claims:

①

AI grouping matches how users mentally organize their tabs

②

Sleeping tabs via Tab Agent frees meaningful browser memory

③

Users find and manage tabs faster with Tab Agent than without

Eval results will be published at /evals after the study completes.

Your tabs,intelligently managed