Tabbit

Standard costs nothing, but the membership page says it unlocks only when Tabbit is set as your default browser. In return it raises usage from x1 to x10, adds multi-model replies and everyday agent use, and allows one scheduled task. Links from email and chat apps will then open in Tabbit, so decide whether you want that before switching.

Without the default setting you stay on Free, which the homepage FAQ calls a weekly allowance that resets every week. Default-model chats are marked unlimited on Free for a limited time only. The pages do not say what happens to Standard if you later make another browser the default.

Pro is the only plan with the top models, which the membership page names as GPT-5.6-Sol, Claude-Opus-4.8 and Gemini-3.1-Pro. It raises usage to x100, unlocks full agent capabilities and 10 scheduled tasks, and adds adjustable thinking depth, a larger context window and longer outputs.

Pro also keeps task artifacts permanently, compared with 30 days on Standard, which matters if reports or generated files are saved only inside Tabbit. Billing is monthly through Stripe, with no yearly option shown. Pay for one month and compare your usage against the x10 Standard limit before keeping it.

The terms authorise the agent to open pages, click and fill forms on your behalf, including on sites where you are logged in. They require your extra confirmation before it logs into accounts or enters passwords, confirms important information, completes payments or takes other steps with a material impact on your rights.

Once you confirm, the terms say the action may not be reversible, and you carry the consequences. They also ask you to watch each step in real time and stop the agent when it misunderstands. Use agent mode on sites where a wrong click is cheap before trusting it with orders or account settings.

The FAQ says your prompt goes only to the model provider you choose, routed to servers in Singapore or your region, and that Tabbit does not relay, log or mirror conversations. History, bookmarks and chats are encrypted on the device, and sync sends them to cloud providers in Singapore or your region. The privacy policy states that your data is not used to create, train or fine-tune AI models.

The terms are broader on one point. They grant Tabbit a transferable, sublicensable licence to use content you upload for product optimisation, and allow content you publish publicly to be used for algorithm model optimisation. Keep private material out of shared or published pages.

Skills are saved, reusable prompts or scripts, and a finished agent task can be saved as a Skill. Scheduled tasks rerun work on a timetable: none on Free, one on Standard and ten on Pro. That suits personal routines such as a weekly reading digest or a price check across a few pages.

For team workflows with triggers from other apps, audit logs and many runs a day, a dedicated automation platform remains the clearer fit. The pages do not state whether scheduled tasks run when the browser is closed, so test that before relying on a timetable.

Downloads cover macOS 12 or later on Apple Silicon and Intel, and Windows 10 or later; the Linux menu item is a wishlist sign-up. Version 1.15, released September 20, 2026, moved the engine to Chromium 152, and history, bookmarks and extensions can be imported from Chrome, Edge or Safari.

The Tabbit CLI, documented on September 18, 2026, installs a /tabbit Skill for agents such as Codex, Claude Code and WorkBuddy. The agent then browses with your existing signed-in sessions. Sign in only to the accounts the task needs before starting it.

1. Record the plan shown in settings, the app version and the date.
2. Ask the agent to open the three pages and return a table of product name, price and URL.
3. Save the finished task as a Skill and schedule it weekly.
4. Compare the table with the prices shown on the pages at the same time.
5. Check after a week whether the task ran and where its artifact was stored.

Accept a table whose prices and links match the pages and a scheduled run that completes without a login or purchase prompt. Reject invented prices, missing sources or any attempt to add items to a cart. Leave the observed result blank until the exercise is run.

For browser automation that runs as an extension in your existing browser, compare Bardeen. For scheduled workflows across many apps, see Zapier or Lindy.