Tabbit

What is Tabbit?

Tabbit is an AI-native browser that integrates AI-driven conversations, web browsing, document organization, and task execution. Users can directly use tabs, screenshots, bookmarks, and local files, allowing the model to provide responses or carry out actions based on the current working context.

It is not a regular extension that can be installed on Chrome; rather, it is a standalone browser. The service in the Chinese market is provided by Beijing KuXun Interactive Technology Co., Ltd. and its affiliated companies, while an international version is available on their official website.

Main functions

• Access websites, conduct traditional searches, and engage in AI conversations through the address bar.
• Cite the current webpage, tag group, screenshot, favorites, and local files.
• Ask questions directly about the selected text or webpage screenshot.
• Switch between multiple domestic AI models on the same interface.
• Use task mode to automatically browse pages, extract information, and fill out forms.
• Save commonly used prompts, scripts, and tasks as tips.
• Generate and execute web scripts in natural language.
• Automatically organize tabs and manage them using a vertical sidebar.
• Save the entire text of a webpage, its images, or selected text, and perform full-text searches.
• Synchronize bookmarks, passwords, conversations, and other browsing data.
• Create a scheduled task and save the file generated by the Agent.

What can the input field on the home page be used for?

Contextual citation

The core feature of Tabbit is allowing the content that is already open or saved in the browser to be incorporated directly into the AI context. Users do not need to copy and paste repeatedly; they can enable the model to process multiple tabs, screenshots, and files at the same time.

• Quote the current tab and summarize its content.
• Cite a tag group in multiple web pages.
• Upload PDFs, tables, presentations, or images.
• Extract a portion of the web page and ask questions related to that content.
• Select the text to explain, translate, or rewrite it.
• Retrieve historical materials from the full-text collection.

Contextual dialogue tutorials

1. Open the web page or document tab that you need to read.
2. Open Chat in the input field of the new tab or in the sidebar.
3. Enter @ and select the current page, tag group, favorites, or local file.
4. Indicate whether you need to summarize, compare, extract, or generate content.
5. Choose the appropriate model and level of thought for the task.
6. Check whether the answer accurately cites the selected sources.
7. Return to the original webpage for verification of numbers, dates, and conclusions.
8. Save useful results as files, favorites, or for future tasks.

Multi-model dialogue

The official website for the Chinese market lists models such as Kimi, GLM, DeepSeek, MiniMax, DouBao, Qwen, and LongCat; the actual list of available models may change depending on the integration strategies in place. The standard version allows multiple models to respond simultaneously, while the professional version offers top-tier models as well as greater context handling capabilities.

• Switch to different models for the same problem.
• Compare the answers of multiple models side by side.
• Select a model based on writing, analysis, and task execution.
• Configure a higher level of thinking intensity and longer input time.
• Unlimited conversations can be carried out using the default model, but this entitlement may change.

Agent task mode

In task mode, it is possible to open web pages independently, read content, perform clicks, fill out forms, and generate structured results. Each task can run within its own tab group, though the user still needs to monitor the key steps.

• Collect and compare product information across different web pages.
• Read the paper and prepare a draft of the research report.
• Process PDF, Word, Excel, PPT, and images.
• Create documents, presentations, or web pages.
• Extract page tables, text, and media resources in bulk.
• Draft and submit repetitive responses in accordance with the rules.
• Prepare the shopping items or form details before final confirmation.

Web Automation Tutorial

1. First, enable the task mode on the test account and on websites with low risk.
2. Specify the goals, the websites that are allowed to be accessed, and the actions that are prohibited.
3. Specify the fields to be extracted and the final output format.
4. It requires suspending confirmation when it comes to payment, publication, deletion, or submission.
5. Observe the tabs opened by the Agent and its execution process.
6. Stop immediately if an error is found in the website, amount, or recipient.
7. Check the final report, forms, and downloaded files.
8. Once verified, high-risk actions are carried out manually.

Agent security boundary

• Do not allow the Agent to automatically complete payments and transfers.
• Do not let it delete files in the cloud, emails, or business data.
• Do not automatically enter your account credentials on unknown websites.
• Conduct individual checks for each item before sending messages or publishing content in bulk.
• It should only be used as a supplement when making decisions related to medicine, law, and investment.
• Changes to the web page may cause the script to click on the wrong location.

Tips and Scripts

This useful feature is designed for saving prompts, scripts, or entire tasks that can be called repeatedly. Users can also use natural language to generate JavaScript or CSS code in order to clean up pages, extract resources, or temporarily change the appearance of a webpage.

• Save frequently used summaries, translations, and writing instructions.
• Convert the completed Agent process into a repetitive task.
• One-click removal of distracting elements on web pages.
• Extract images or links from a page in bulk.
• Share or use the useful tips provided by the community.
• Refreshing the page will remove the temporary style changes.

Checks before using community tips

• First, read the script and check the permissions for the request.
• Do not run it on online banking, payment, or backend management pages.
• Do not execute code that requires uploading cookies or tokens.
• Use the test page to observe network requests and the scope of changes.
• Re-validate the script’s functionality after the webpage is updated.

Tab and bookmark management

Tabbit makes use of a vertical tag bar and AI-powered sorting to organize large numbers of pages. Its bookmarking function saves the entire content of web pages, not just the links; it also allows for the creation of summaries, as well as the saving of selected text or images, which can later be retrieved through full-text search and RAG technology.

• Organize the open tabs by topic with one click.
• Save the complete tag set as project data.
• Collapse the vertical bar to get a larger viewing area.
• Save the full text of the webpage as a backup in case the original page becomes unavailable.
• Generate a summary for collection purposes to enable quick review.
• Organize ungrouped collections and search by topic.

Data migration and multi-device synchronization

The official website states that it is possible to transfer bookmarks, browsing history, extensions, and custom settings, as well as synchronize bookmarks, passwords, and conversations. The membership page lists macOS, Windows, iOS, Android, and HarmonyOS; however, the range of functions available may vary across different platforms.

Member packages and prices

As of August 20, 2026, the member page for the Chinese market offers an experimental version, a standard version, and a professional version. The amount of usage is displayed as a relative percentage rather than in terms of a fixed number of tokens; the actual consumption depends on the model and the type of task.

Main differences between packages

The Professional version of the blog also mentions that it is possible to purchase usage reset vouchers as needed; upon use, the remaining usage for the current period is restored, without altering the original seven-day cycle. Prices, models, exemptions, and usage multipliers may change, so it is necessary to refer to the purchase page on the client side.

Key points of the privacy policy

The current privacy policy was updated on June 8, 2026, and came into effect on June 15. It applies to both desktop and mobile versions of the application, and specifies that AI services process text, files, images, audio, commands, as well as a history of questions and answers that users enter voluntarily.

• To register an account in China, a mobile phone number is required at minimum.
• The search function handles keywords, images, voice, and search behavior.
• Some search records on the PC version are saved locally and can be deleted.
• The content copied to the clipboard in normal mode is recognized only locally.
• Collections and conversations can be uploaded to a server to achieve cross-device synchronization.
• The AI feature shares the necessary instructions with the selected model service partners.
• The policy clearly states that such data is not used for model training or fine-tuning.
• Users can manage history, cookies, permissions, and request to delete their account.

Incognito browsing does not mean leaving no trace at all.

When exiting or restarting in incognito mode, the local search history, browsing history, Cookies, and cache are deleted, but the service still collects the device information and logs necessary for its operation. It is not possible to hide the activities that are visible to network operators, websites, and legitimate security logs.

Suggestions for privacy protection

• Do not include ID cards, medical records, contracts, and customer data in the AI context.
• Close unrelated and sensitive pages before citing the entire tag group.
• Different websites use separate passwords and enable multi-factor authentication.
• Regularly clean up conversations, search history, favorites, and task outputs.
• Turn off permissions for unnecessary microphones, location, photo albums, and notifications.
• Before using a third-party model, understand its data processing rules.
• Delete the account and personal information before sharing tips, web pages, or reports.

What use cases are suitable?

• Researchers read and compare multiple web pages or papers at the same time.
• Content creators organize materials, outlines, and the context for writing.
• Product managers gather information on the features and prices of competing products.
• Operations staff handle repetitive customer service tasks and data entry.
• Students translate, explain, and summarize the course materials.
• Analysts extract structured information from web pages and tables.
• Powerful browsers create numerous tags and save entire texts.
• Users who wish to create lightweight web scripts using natural language.

Tasks that are not suitable for completion directly

• Unsupervised payments, ordering, and financial transactions.
• Delete files, emails, or online business records in bulk.
• Process highly confidential, regulated, or unmasked data.
• Guarantee completely accurate papers, legal, or investment research.
• Bypassing website verification codes, access controls, and terms of service.
• Stable alternatives to professional RPA and proven enterprise automation solutions.
• A working environment with the native Linux desktop version is required.

Product advantages

• AI conversations are closely integrated with the browser context.
• It supports cross-tab referencing, screenshotting, bookmarking, and local file references.
• Multi-model switching is suitable for comparing different responses.
• Agents can perform complex tasks in independent tag groups.
• A clever trick that allows prompts, scripts, and tasks to be reused.
• Vertical tags and smart grouping are suitable for intensive browsing.
• Saving the entire text allows for easier long-term retrieval compared to just keeping the link.
• The standard version can be unlocked for free by setting it as the default browser.
• Synchronization between desktop and mobile devices is fairly complete.
• The privacy policy clearly states that it is not to be used for model training or fine-tuning.

Usage restrictions and precautions

• The agent may misread the page, click on the wrong control, or submit repeatedly.
• Changes to the website interface can cause scripts and automations to stop working.
• Summaries, tables, and research reports generated by AI may contain errors.
• The availability of different models varies depending on the region and package.
• The quantities are only specified as multiples, making it difficult to estimate the exact number of tasks.
• The default model can be used unlimited times as part of the current benefits; this may change in the future.
• Cross-device synchronization saves some favorites, passwords, and conversations on the server.
• Third-party websites and models have their own separate privacy and service policies.
• Scripts for natural language generation can pose risks to website and account security.
• The functions on mobile devices are not exactly the same as those on desktop devices.
• There is no public version of Linux for desktop installation.
• High-risk actions must require manual confirmation.

GitHub and the open-source status

A real-time search did not yield any repository containing the complete browser source code, nor any open-source license that has been confirmed by Tabbit’s official website. Tabbit should be considered a proprietary commercial browser; the ability to run user scripts and compatible extensions does not mean that the browser itself is open source.

Basic information

Recommendation score

4.6 / 5. Tabbit integrates multi-model capabilities, browsing context, Agent tasks, tips, and tag management into a cohesive desktop workflow; its free standard version is attractive as well. However, automatic web execution, cloud synchronization, and AI-related errors still require user oversight.

Frequently Asked Questions

Is Tabbit free?

The trial version is free; by setting it as the default browser, one can unlock the standard version with a higher data limit for free. The professional version costs 39.60 yuan per month.

Is Tabbit a Chrome extension?

No. It is an AI browser that can be installed independently, but it allows for the transfer of browsing data and compatibility with certain extension ecosystems.

Which operating systems are supported?

The desktop version is compatible with macOS 12 and later, as well as Windows 10 and later; the member page also indicates synchronization across iOS, Android, and HarmonyOS devices.

Can web pages be operated automatically?

Yes. The task mode allows for opening pages, extracting information, and filling out forms, but actions such as payment, submission, deletion, and publication require manual confirmation.

Will dialogue be used to train the model?

The current privacy policy clearly states that such data will not be used for model training or fine-tuning. AI services still need to process the input data and may send the necessary information to the selected model partners.

Does incognito mode completely hide activities?

No. It primarily removes local browsing history, cookies, and cache; however, logs related to essential devices and services may still be collected.

Does Tabbit provide an API?

No public API documentation available for developers was found. Built-in agents, scripts, and extension capabilities do not equate to external APIs.

Is Tabbit open source?

It is not open source; no complete source code repository or open source license confirmed on the official website was found.