Product changes, launches, workflow upgrades, and trust improvements from the TabMate build.
This page is the running record for what changed, when it changed, and which parts of the product got sharper across access, reliability, control, and day-to-day research workflow.
You can now enable Agent Mode to let TabMate interact and perform tasks on the page on your behalf.
This addition gives TabMate truly agentic behavior and it can now safely be called an assistant.
TabMate will now keep its notes up-to-date with the latest question, leading to better understand of your current goals.
Previously, notes would lag, leading to a loss of performance - both in speed and accuracy. Now, TabMate will take notes live.
Approved and rejected memory candidates will no longer reappear when TabMate generates them.
Now, TabMate is aware of what you saved and rejected and won't surface them again to reduce the review noise you previously had.
For pages that have very less text content or when TabMate cannot parse the page, it will ask you to take a screenshot.
This gives TabMate another way to see the page when it cannot read it, thus allowing you to continue across sites that would previously not work.
TabMate is now 2x faster during long-context questions and 1.5x cheaper.
Extend your use and get the most out of your credits with this release.
You can now buy top-up credits whenever you run out of usage limits. Purchasing a top-up also resets your daily ask and token limits.
These credits rollover and get used when you continue once daily or monthly limits are reached.
Implemented token caching to prevent rebuilding the entire context at every ask.
This reduces response times for common, repeated asks and cached tokens help save credits.
You can now link your account to your Google account.
Saves you the hassle of remembering multiple passwords.
TabMate now understands your work scope and adapts as your goals change.
Users can select their personas. Starter playbooks change based on the persona. Student users will be required to verify their age.
We are now listed on the Firefox Add-ons Web Store for extensions.
Users can now install TabMate with lesser friction and manual workflow.
We are now listed on the Chrome Web Store for extensions.
Users can now install TabMate with lesser friction and manual workflow.
On long pages, TabMate may hint you to scroll to relevant sections if it does not find enough information for the ask in its current page view.
This ensures that the responses you get are not based on partial evidences but grounded in relevant page content.
You can now delete your account from within TabMate instead of relying on a manual support path.
Deletion disables the account immediately, clears TabMate’s local extension data on the device, and schedules associated server-side data for permanent removal.
Conversations, memories, prompts, and workspaces can now be deleted directly from within the product.
That gives you tighter control over what stays saved, what gets cleared, and how much old context you want to carry forward.
Conversation compression is now part of the workflow, so longer research threads stay usable instead of gradually collapsing under their own weight.
The goal is not to keep every word forever. The goal is to preserve continuity without forcing you to start cold again.
TabMate now does a better job of surfacing memory candidates and keeping recurring context available across sessions.
Useful details are less likely to disappear into old threads and more likely to stay attached to the work they belong to.
Working context such as goals, stage, and related notes is now tracked more explicitly inside the product.
That makes the side panel feel less like a disconnected chat box and more like an ongoing workspace with a live thread of work.
Streaming responses can now be interrupted, and response flows can be rerun more cleanly when you want another pass.
That sounds small, but it matters. A tool like this should behave like something you can work with, not just wait on.
TabMate now applies stronger content sanitization, forwarding controls, and request-side limits around asks.
That includes more defensive handling of browser-derived content, stored inputs, and certain unsafe outputs before they turn into avoidable mess.
Page evidence now carries forward more reliably through the side-panel workflow.
Pinned excerpts, cached context, and related research state are better aligned with recurring work instead of being treated like one-off asks.
You can now save excerpts, pins, evidence, responses, and custom notes and instructions as memory that TabMate can reuse later.
Pinned excerpts still help in the moment, but longer workflows need context that survives beyond a single page or short session.
