Meet Tab Sage: Spyxpo’s Private, Local-First AI Browser With No Subscription
This Free AI Browser Runs Entirely on Your Computer and Lets You Remove Any Part of a Website
Tab Sage combines private, on-device AI with multiple search engines, webpage summarization, content filtering, built-in ad blocking, open-source extensions, and an unusual tool that lets users remove unwanted elements from websites.
View the official Tab Sage project page
The next major browser may not come from Google, Microsoft, or Apple.
Tab Sage is a privacy-focused desktop browser being developed by Spyxpo. Instead of treating artificial intelligence as a remote service, it runs its language model directly on the user’s computer.
This allows the browser to summarize webpages, answer questions about open content, extract important information, and support page-aware AI conversations without sending the browsing context to a cloud-based language model.
Tab Sage also provides something many AI browsers overlook: direct control over the webpage itself. Users can block advertisements from the top toolbar, temporarily disable the blocker when necessary, and activate an element cutout tool to remove unwanted sections from the screen.
The browser is currently listed as alpha software, with version 1.0.0 in development.
The AI Runs on Your Computer, Not Someone Else’s Server
Most AI-powered browsing tools work by sending text from the current webpage to an external server. A cloud-based model processes the content and returns a response.
Tab Sage takes a different approach.
Its bundled Llama 3.2 3B Instruct model runs locally through llama.cpp. The model uses the Q4_K_M GGUF format, which is designed to make local AI processing more practical on consumer hardware.
When users ask Tab Sage to summarize an article, explain a section, extract a date, or answer a question about the current page, the processing takes place on their device.
Browsing history, AI conversations, page context, and webpage content remain locally managed instead of being routinely submitted to an external AI provider. Once the model has been downloaded, its AI features can continue working without an internet connection.
This does not eliminate every privacy risk associated with browsing the web. Websites can still collect information through their own systems, and search requests must still reach the selected search provider.
The important distinction is that Tab Sage does not require users to send webpage content to a cloud language model whenever they activate an AI feature.
Summarize Almost Any Page Without Copying It Elsewhere
Long webpages are often filled with introductions, repeated explanations, navigation elements, and information that may not be relevant to the reader’s immediate goal.
Tab Sage can summarize the current page directly within the browser.
Someone reading an article can request its main arguments. A person reviewing technical documentation can ask for the most important steps. A researcher can extract names, dates, links, statistics, deadlines, and other significant details without manually scanning the entire page.
The browser’s contextual question-and-answer feature also lets users ask direct questions about the content they are viewing. Responses are grounded in the current page rather than generated as general answers with no connection to the source material.
The result is a browser that treats AI as a practical reading tool rather than a separate destination.
Multiple Search Engines Instead of One Locked Default
Tab Sage is not tied to a single search provider.
Users can choose between search engines such as Google, Brave Search, and DuckDuckGo. They can change the default provider according to their preferences and add custom search engines when the built-in options are not enough.
Custom search support can be useful for more than ordinary web searches. Users could configure direct searches for documentation websites, academic databases, knowledge bases, company portals, product catalogs, or any other service that accepts search queries through a URL.
Search therefore remains under the user’s control rather than becoming a permanent decision made by the browser developer.
The Ad Blocker Is Built Into the Top Bar
Tab Sage includes a built-in advertisement and tracker blocker.
Users do not need to install a separate browser extension before blocking common advertising resources. The blocker can be controlled from the browser’s top toolbar and enabled or disabled whenever necessary.
That flexibility matters.
Some websites refuse to display content when an advertisement blocker is detected. Others may stop working because a required script has been caught by an overly broad blocking rule.
Instead of forcing users to search through multiple settings menus, Tab Sage lets them change the blocker’s status for the current page directly from the browser interface.
Its extension API can also allow extensions to enable or disable the blocker for an individual tab without requiring the page to reload.
This creates a more practical system than a blocker that must either remain active everywhere or be disabled completely.
Do Not Like Part of a Website? Cut It Out
The element cutout tool may be one of Tab Sage’s most noticeable features.
Users can activate cutout mode through a scissors button in the toolbar and click an unwanted section of a webpage to remove it from the screen.
Examples could include:
- A newsletter signup covering part of an article
- A distracting sidebar
- A floating promotional banner
- An oversized header
- A recommendation panel
- A sticky element that follows the reader down the page
- Any other visible section the user does not want to see
This is different from blocking a network request. It gives users direct control over the visible webpage after it has loaded.
The feature is also available to extension developers through the permission-based tabsage.cutout API. Extensions can activate or exit cutout mode, while ordinary users can access the same capability through the toolbar’s scissors button.
Multi-Member AI Goes Beyond a Single Chatbot
Most browser assistants provide one conversation with a single general-purpose AI assistant.
Tab Sage supports multi-member AI conversations, allowing several configured AI participants to contribute to the same workflow.
Each member can be assigned a different responsibility. One could summarize the current page, another could extract factual claims, and a third could identify weaknesses, missing context, or questions that require additional research.
This could make Tab Sage more useful for structured analysis than a single assistant expected to complete every task by itself.
Because the underlying model runs locally, these workflows remain connected to the browser’s device-first AI architecture rather than depending on a separate cloud model for each participant.
Built-In NSFW Protection
Tab Sage includes an NSFW content filter that can block adult websites and other selected categories.
This provides an additional layer of control before unwanted material appears. It may be useful on shared devices, family computers, educational systems, workplace machines, or personal setups where certain categories of websites should be restricted.
Content filtering should not be treated as a perfect replacement for supervision or dedicated network-level controls. Website classifications can be incomplete, and new domains appear constantly.
It is better understood as an additional browser-level protection feature.
Combined with local AI, incognito browsing, tracker blocking, and the stated absence of usage telemetry, the filter forms part of Tab Sage’s broader focus on privacy and user control.
No Telemetry Means No Browser Usage Analytics
Many applications describe themselves as private while still collecting product analytics, diagnostic events, feature-usage information, and behavioral data.
Tab Sage states that it does not collect, track, or transmit browsing or usage telemetry.
Information such as browsing history, AI conversations, bookmarks, settings, and page context is intended to remain on the device. Sensitive credentials are protected through the operating system’s native security facilities.
This does not mean every website visited through Tab Sage becomes private. Websites, search engines, internet service providers, and other online services may still receive information according to their own policies and technical systems.
It means the browser itself is not designed around collecting a secondary stream of behavioral analytics from its users.
Password Vault, Autofill, Bookmarks, and Searchable History
Tab Sage is more than an AI interface wrapped around a basic browser.
It includes many of the everyday features needed for regular browsing, including bookmark management, searchable history, tab search, a visual tab overview, password storage, payment-detail autofill, dark mode, and incognito browsing.
Saved credentials are secured through the operating system’s native protection system rather than stored as ordinary, unprotected application data.
Incognito tabs can exist beside regular tabs in the same window. Pages opened privately are not added to normal browsing history, allowing users to move between standard and private sessions without managing separate windows.
The browser is also designed to let users search previously visited pages and quickly locate open tabs by title or URL.
Developers Can Build Extensions That Use the Local AI
The Tab Sage browser itself is not open source, but its extension ecosystem is.
The official Tab Sage Extensions repository is public and contains the source code for extensions available through the browser’s extension catalog.
Users can inspect an extension’s code before installing it. There is no separate extension-store account system, and the repository states that Tab Sage fetches extension catalog content only while the Extensions page is open.
Submitted extensions are reviewed before being added to the official registry.
For developers, one of the most important features is the tabsage extension API.
Extensions can request permission to access capabilities such as:
- The on-device AI model
- Per-extension local storage
- Current and open tab information
- Tab opening, activation, reloading, and closing
- Browser notifications
- Custom dialogs
- Ad-blocker controls
- Element cutout mode
- Clipboard access
- File downloads
- Toolbar badges
- Cross-tab extension messaging
- Page text, HTML, selections, and metadata
- Extension runtime information
- Keyboard shortcuts
Permissions must be declared in the extension’s manifest. If an extension has not requested access to an API group, that group is not exposed to it.
The AI API is especially significant. Extension developers can send prompts to the same local model that powers Tab Sage’s main assistant.
This makes it possible to build page summarizers, research utilities, writing tools, accessibility features, productivity systems, and page-aware assistants without requiring every extension to connect to an external AI provider.
Developers can explore the repository here:
Extension Source Code Is Public and Reviewable
Every extension in the official repository has its own folder containing a manifest and the scripts, styles, documentation, and other files it needs.
The repository includes development tools, a command-line interface, and a reusable user-interface kit.
Reference extensions demonstrate features such as reading-time calculation, word counting, sticky notes, local-AI page summarization, page-aware chat, clipboard tools, and tab management.
The system is designed around relatively simple extension packages. Extensions can remain free of external dependencies, and the files inside each folder are the files executed by the browser.
This public model gives developers examples they can learn from while giving users an opportunity to examine the code behind extensions before deciding whether to trust them.
There Is Even a UI Kit for Browser Extensions
Extension developers often have to build interfaces from scratch or import large frontend frameworks.
Tab Sage provides an open-source UI kit with reusable components designed to match the browser’s visual style.
It includes components for buttons, inputs, cards, switches, selection menus, context menus, toolbar popups, launchers, modals, and notifications.
The kit follows the system’s light or dark appearance and does not require a network connection or runtime package dependency.
This gives independently developed extensions a better chance of feeling like a natural part of the browser rather than unrelated scripts placed on top of a webpage.
Your Computer Must Be Capable of Running It
Local AI comes with a cost: the user’s device must perform the computation.
Tab Sage recommends at least 8 GB of RAM and a multi-core processor. Performance will vary depending on the processor, available memory, operating system, and number of other applications running at the same time.
To reduce the chance of users downloading a model their computer cannot run comfortably, Tab Sage includes a hardware benchmark and device-capability test.
The browser evaluates the computer during the initial setup process. The benchmark can also be run again from the Settings menu.
On devices below the recommended threshold, the model download is blocked by default, although users can choose to continue manually.
This is a more responsible approach than allowing a large download and revealing the performance problem only after installation.
It Is Free, but the Main Browser Is Not Open Source
Tab Sage is intended to remain free to use.
Local AI questions are not charged by token, and webpage summaries do not require a metered cloud subscription because the processing takes place on the user’s own hardware.
However, there is an important distinction between being free and being open source.
The main Tab Sage browser is free but not open source. Its extension registry, reference extensions, developer tools, APIs, and extension UI resources are publicly available through the open-source extension repository.
This structure allows Spyxpo to retain control over the main application while still providing an open development layer around it.
The Most Interesting Part Is Not Just the AI
Calling Tab Sage an AI browser does not fully explain what makes it different.
Its more compelling idea is that several important controls are placed back into the user’s hands.
Users can choose their preferred search engine or add a custom one. They can turn the advertisement blocker on or off from the top toolbar. They can remove distracting sections from the current webpage. They can inspect the source code of extensions, and developers can build tools that use the same local AI model as the browser.
Page content does not need to be submitted to a cloud model for basic AI assistance.
Tab Sage is still in development, so its final performance, compatibility, and everyday reliability cannot yet be evaluated as though it were a finished browser.
Its planned feature set, however, presents a clear vision of what an AI browser could look like when privacy and user control are treated as part of the core architecture rather than optional settings.
For years, the browser wars have focused on speed, synchronization, and ecosystem lock-in.
Tab Sage is betting that the next competition will be about who controls the AI, who controls the data, and who controls what appears on the screen.