AI Group Tabs vs Tabbit

The agent can read open tabs, PDFs, bookmarks, HTML elements, screenshots, highlights and local files. Mention anything you have on screen — '@' a tab, a file or a highlighted element — and the agent picks it up as context.

Access most major models from inside the browser, with new versions typically available within about 12 hours of launch. The active model is displayed at the top of each chat so you always know which one is answering.

A large library of pre-built Skills covers popular sites and everyday tasks — pulling highlights from long videos, exporting a transcript to a knowledge base, summarising a research feed, or auditing subscriptions from email receipts. You can also package your own recurring prompts into Skills.

Research, operator, writing and analysis roles can be loaded with the right Skills and run in parallel for bigger jobs, so a single request can fan out across several specialised agents instead of one generalist chat.