The things a browser makes you do by hand every single day, and the part of the extension that stops each one. Nothing here needs an account, a server, or more discipline than you already have.
The tab you want is a four-pixel icon in a strip that stopped being readable an hour ago. You reopen pages you already have open, twice a day.
You drag the work tabs into a group, name it, colour it. By the afternoon there are nine loose ones again, and tomorrow you do it a third time.
A bookmark, something in your history, a note, a screenshot: each one lives on its own page, and going to fetch it takes you off the page you were on.
Open a link, find a tab, jump to a group, close six of them: every one of those is a trip to the mouse and back, a few hundred times a day.
A rule is a list of URL fragments and the group they belong to. Any tab whose address contains one of them lands there, with the name and colour you gave it. What has no rule is grouped by domain, subdomain, IP address or local file.
A group nobody has touched folds itself when its timer runs out, the duplicate counter clears them in one click, and a whole group can be backed up and reopened days later exactly as it was.
Bookmarks, history, recently closed tabs, the reading list, downloads, notes, the screenshot gallery and the assistant are nine views of one side panel, beside the page rather than instead of it.
f draws a letter on every link, o opens a search box over the page with a prefix per source, and the agent takes the job in plain language. Every binding is editable.
Every other thing the extension does belongs to one of the panels below. Each one lists what it actually does: no capability is named here that you cannot find in the side panel today.
The assistant lives in the side panel and holds fifty-eight tools onto the browser itself: tabs, groups, rules, bookmarks, themes, snippets and site shortcuts. Ask in plain language and it runs them.
Grouping is decided by rules, not by guesswork: a rule names the addresses it owns, the group they land in, and its colour. Everything else here exists to keep that list of groups short.
Two things that work together: a record of the time each site takes, and rules that stop a site once it has had enough of your day. Everything is kept on this device.
Two ways in: a floating search box over any page, and letter labels drawn on the page itself. Neither of them needs the mouse, and neither is fixed. Every binding on this page is editable.
Press f and every link, button and field on the page wears one or two letters. Type them and it opens; hold Ctrl and it opens in a background tab. It works on this one toogj, and in any text field on it two dollar signs open the snippets you use most.
Press o and the omnibar floats over whatever you are reading. Each prefix searches a different thing, and one of them hands the sentence straight to the agent. Prefer the address bar? Type find there and you get the same search without leaving it.
b: bookmarks, h: history, c: recently closed, lnt: notes, limg: screenshots, rl: rules, bgr: backups, lai: past AI conversations. Type @ and it lists every prefix, so you never have to remember one.
dg: closes whole groups, dt: closes several tabs at once, ts: puts the ones you pick side by side, f: finds text inside the page, ar: reads it aloud, cr: writes a grouping rule, qaia: gives the job to the agent.
Press f and every link, button and field wears a one or two letter label. Ctrl opens it in a background tab, Shift in a new window, and cf copies the link instead of following it.
lnt: noteslimg: screenshotsrl: rulesbgr: backupsReader mode takes the page down to its text. The floating player then reads it, highlighting the paragraph and the word as it goes, at the voice, speed and pitch you choose.
The paragraph and the word being spoken are marked as the voice moves through them, and you set how solid that highlight looks.
One key strips the article out of its layout. Dark, sepia, grayscale or light can also be applied to any site as it is, for one tab or for every tab at once.
Language, speed, pitch and volume; start it with ar from the keyboard, from the omnibar, or on the assistant’s own answers.
The reader takes the article out of its layout: no menus, no banners, no floating boxes. The player then reads it, and the paragraph it is on is marked, along with the word being spoken, at an opacity you set yourself.
Language, speed, pitch and volume are yours to set, and the two marks keep the colours of whichever theme the extension is wearing.
A picture-in-picture button appears on YouTube, Shorts, TikTok and ordinary HTML5 players. The whole page can float too, and the loop button lives in the player where you would expect it.
Send a video out with wv, or let it leave on its own: as floats it once you scroll past it, ah when you switch away from the tab.
Added to the YouTube player and to Shorts, so a track, a lesson or a background loop repeats without touching anything.
wp puts the current page in a document picture-in-picture window and we in a clean popup, both also reachable from the omnibar with wp:, we: and wv:, and from the right-click menu.
Something the browser has never been good at: keeping a picture of what you were looking at, and putting a set of tabs somewhere safe until you need them again.
docs.astro.build (full page)
IW-214 · area
Pricing table
Invoice · June
Not just the visible screen. The whole page, scrolled and stitched into one image up to 30,000 pixels tall; the same page split into parts when that is easier to read; an area you draw yourself; or every tab of a group in one go.
A backup is a whole group closed and kept: its tabs, their order and their titles. It stays in the list with a B on it, and one click opens every one of them again.
The side panel holds today; the full page holds the rest. Time per site and per category, and rules that stop a site once it has had enough of your day. All of it recorded on this device, and none of it sent anywhere.
One box per site, folded, with the time it has taken. Unfold it and you get visits, average time per visit, its share of the day, and the three rules that govern it.
When one runs out the site is actually blocked, on a screen that says why and when it lifts. A password can be required before a rule is switched off, and "five more minutes" is recorded as a snooze.
Time browsing, focus ratio, peak hour, busiest weekday, stretches, streak and top category, with the hour-by-hour grid, the per-category split, and every visit on a timeline.
Chrome keeps five of these lists in pages you have to go and open. Here they are five views of one panel, with a search that filters as you type and an action on every row.
Search across every folder, create, rename, delete and drag things around, import and export, and turn a single bookmark or an entire folder into a grouping rule.
A scan visits each saved address and reports the ones that fail, sorted by what went wrong: 4xx, 5xx, timeout or a connection error. Filter to one kind and delete them together.
History by day with a date filter, a recently closed list that brings back a whole window in one click, and downloads grouped by date and filtered by status: complete, in progress, interrupted.
Type a short trigger and the full text appears, formatted or plain, with variables filled in on the spot, in any text field on any site.
Give one a default with {{name|there}} and it fills itself in unless you say otherwise. The editor checks the syntax while you write it, so a broken snippet never gets saved.
A rich text editor for bold, lists and links; add # to the end of the trigger and the same snippet pastes with no formatting at all.
In any text field, including Google Docs, Word Online and WhatsApp, where ordinary expanders give up.
;pr# plain text, no formatting$$ opens your five most used{{date|today}} default valueEvery row below is a job people usually install a separate extension for. They share one side panel, one settings page, one set of shortcuts and one permission prompt, and none of them phones home.
Each one with its own permissions, its own settings page and its own toolbar button, and several of them keeping a copy of your browsing on a server you do not run.
A grouper, a session saver and a tree-tabs sidebar: three UIs, three sets of shortcuts, three ideas of what a group is.
A suspender that replaces your tab with its own placeholder page, and loses the tab if it is ever uninstalled.
A chat panel that can read the page and nothing else: it cannot close a tab, make a group or write a rule.
One extension for a command palette, another for link hints, and a fight over which one owns which key.
A notes app with an account, a bookmark manager with a subscription, and Chrome’s own pages for the rest.
A tracker that uploads your browsing to somebody’s server, and a blocker that a new tab gets around.
Five one-trick extensions, each with its own permissions and its own toolbar button.
One Manifest V3 extension, one side panel, one place where every shortcut is defined, and storage that never leaves this device.
Rules by URL fragment; grouping by domain, subdomain and IP; auto-collapse timers; group backup and restore; recently closed and saved sessions.
Chrome’s own tab discarding after an hour idle: the tab keeps its title and its place, and wakes up on click.
Fifty-eight tools onto the browser, plus summaries, saved conversations, scheduled questions, and Chrome’s on-device model when the quota runs out.
One floating omnibar with a prefix per source, letter labels on every link, and every binding editable in one place.
Notes, checklists and Kanban boards, the bookmark tree, history, recently closed, the reading list and downloads, all in the same panel, all local.
Time per site and category on this device only, daily and weekly caps, opening hours, and a block screen that can ask for a password.
Full-page and area capture with a local gallery, on-device OCR, QR reader and generator, a cookie editor, and a screen colour picker.
Three ways in, and none of them needs the mouse: a search box that floats over the page, letter labels drawn on the links themselves, and the browser-level shortcuts. Every key here can be reassigned.
One box, and a prefix per source. Type @ if you cannot remember one and it lists them all.
One or two letters, typed on the page itself. With hints on, every link and field wears one.
The four Chrome-level combinations, and the page keys that open one side panel view or another.
Chrome will tell you this extension asks for 24 permissions. That is a fair thing to be suspicious about, so here is what each group of them is for. Nothing is uploaded, because there is nowhere to upload it to.
Groups, rules, notes, screenshots, time records and Pomodoro history live in your browser’s own storage. There is no server to send them to.
The Google AI Studio key you paste is kept in local storage and used to talk to Google directly. Chrome’s on-device model needs no key and no connection at all.
No tracking pixel, no fingerprinting, no telemetry endpoint, and nothing that has to be turned off in settings.
Everything you need to know about privacy, performance, and architecture.
Intelligent Workspace is completely free. There are no paywalled features, no subscriptions, and no advertisements. Optional donations help support ongoing maintenance.
No. There is no login, no email capture, and no authentication server. All your settings and data remain on your device.
You paste your own free API key from Google AI Studio; it is kept in local browser storage and used only when you prompt the assistant, going straight to Google. You can store more than one, and when they run out of quota the extension falls back to Chrome's on-device model, which needs no key and works offline.
None. Groups, rules, notes, screenshots, time records and Pomodoro history are stored by your own browser. The only outbound traffic is the request you make to Google when you prompt the assistant, and even that stops if you use the on-device model.
Chrome and the Chromium browsers that implement the Side Panel and Tab Groups APIs: Brave, Edge, Opera, Vivaldi. Firefox and Safari do not have those APIs, so it cannot run there. Chrome's on-device AI model is a Chrome feature and needs a machine that qualifies for it; everything else works without it.
Yes. The whole repository is on GitHub so that anyone can audit it and open a pull request. The licence is proprietary rather than open source: you may read it, build it for your own private use, and propose fixes, but not republish it or ship a derivative.
No investors, no bloated roadmap. Built with craft for users who demand peak performance and complete control over their browser.
“We built Intelligent Workspace because we believe your browser should be your most powerful workstation, not your biggest distraction.”
Special thanks to Flor Chávez for her relentless dedication to intuitive design and rigorous QA.
Intelligent Workspace is free, private, and ad-free. If it gives you your focus and afternoons back, a one-off donation keeps development active.
Payments processed securely by Stripe. We never see or store your payment details.