People have been experimenting with AI tools that can do more than answer questions for some time.
Many of those tools still require switching to a separate window or a specialized browser. That setup works for some tasks, but it adds friction when the work already lives in ordinary browser tabs, from logged-in dashboards to forms and internal sites.
Perplexity has been building Computer as an agent that can handle multi-step work across tools.
The company has now made a version of that agent available inside the Chromium browsers most people already use.
The company said Computer works directly in those browsers through its official extension.
A person can open a side panel next to the page they are viewing, describe a task, and watch the agent operate across open tabs and other tools. The same extension is listed for Chrome, Edge, Brave, Arc, and other browsers that can install from the Chrome Web Store.
Tasks can also be started from the main Perplexity site. The company has not required users to move to its own Comet browser for this capability. Comet already includes similar agent features without needing the extension.
Before this update, starting in late October, Computer could control a Chromium browser only after a person began the task inside the Perplexity Mac or Windows desktop app.
The app then used the extension to act in the person's already-signed-in tabs.
Now, as shown in the November 20 post and its demo, a person can hand off the task directly from the browser's side panel and watch Computer work across open tabs and tools without first switching to the desktop app.
The underlying capability stays the same.
It still covers user-initiated form filling, multi-tab work, and access to logged-in sites through the same extension. The main change is a more browser-native entry point and the ability to see the work in real time while it runs.
A short video posted with the announcement shows two examples.
In one, the agent opens a supplier registration page, fills in company details across the required fields, and stops before submitting. In another it checks the same product launch on three different ad platforms, compares the landing page URLs against the expected one, and notes which campaign points to a different address.
It leaves the tabs open for review rather than closing them.
The extension listing and the company both state that Computer only acts after a person starts a task and only inside the browser they have pointed it at.
It does not collect browsing data in the background. It can work with sites a person is already signed into, such as email, billing dashboards, or other tools, as long as the site allows it and any extra authentication steps are handled.
Users stay in control of when a task begins.
The extension itself can be added at no charge.
Running Computer still uses the same plan and credit system that applies when the agent works elsewhere. The company has not announced separate pricing for the browser version.
For people who already have access to Computer, the change mainly removes the need to start every browser task from the desktop app.
They can stay in the tabs they already have open and hand off work from the side panel. The agent then creates or uses additional tabs as needed and reports progress.



























































































































































































































































































































































































