What happened
Perplexity’s Chrome Web Store listing describes an extension named Perplexity for use with the Perplexity Computer desktop app. Its stated purpose is to let Computer “drive your browser”: open sites, read pages, fill forms, and finish tasks in tabs the user already uses. The extension is listed in the Tools category, is version 0.0.6, has a listed size of 128 KiB, and was updated on October 25, 2026.
The important capability is access to a local, user-selected browser session. Perplexity says that, after the extension is installed, Computer can act inside sites where the user is already signed in. Its examples include checking a LinkedIn feed, retrieving invoices from a billing dashboard, opening CNN and summarizing top stories, completing applications or booking flows, and comparing listings or combining information from several pages. The listing explicitly frames the scope as email, feeds, dashboards, and internal tools, not only the open web.
The setup is specific. A user must first install Perplexity Computer on a Mac or Windows PC, then add the browser extension. On Mac, the selected browser is set under Settings → Connectors; on Windows, it is selected under Settings → Permissions. The browser selected in the desktop app must match the browser in which the extension was added. The listing names Chrome, Edge, Brave, Arc, and other Chromium browsers as compatible. Users of Perplexity’s Comet browser do not need the extension because the same capabilities are built into that browser.
Perplexity describes two boundaries on operation. Computer uses the browser only during a task started by the user, rather than collecting browsing data in the background. It also acts only in the browser that the user pointed it to. The Web Store privacy disclosure says Perplexity handles authentication information and website content. It further declares that the data is not sold to third parties outside approved use cases, is not used or transferred for unrelated purposes, and is not used to determine creditworthiness or for lending.
This is a product release rather than a new model or API. Its practical change is the connection between an agent running in a desktop app and a real local browser profile: the place where many teams already hold authenticated sessions, operational dashboards, and browser-only workflows.
What the archive adds
This is a new fact on a known Perplexity Computer story. Perplexity Computer now runs fully local on Apple silicon Macs, with model and agent on-device examined local execution and optional cloud escalation. The new extension adds a separate capability boundary: Computer can now act through a selected Chromium browser session.
Trend signal
This supports Agent-ready software exposes capabilities. The extension presents browser control as an explicit, installable and selected capability rather than an implicit property of a chat interface. It also supports Agent security moves to runtime boundaries: the relevant boundary is the chosen local browser and its authenticated session, not merely an instruction telling an agent what not to do.
Why it matters
For teams building with AI, authenticated browser access is often the difference between a demonstration and useful automation. Many operational systems have no convenient API for every workflow, while their web interfaces contain the live state needed to do work. A browser-capable desktop agent can therefore span inboxes, vendor portals, dashboards, internal applications, and multi-tab research without requiring each system to be integrated separately.
That convenience changes the engineering problem. The question is no longer only whether the agent can navigate a page. Teams need to decide which browser profile it may use, what credentials are exposed through that profile, which form submissions need review, and how a user can recognize an action before it has external consequences. Perplexity’s stated task-start and selected-browser constraints are useful product boundaries, but they do not substitute for a team’s own authorization and review design.
Limits and uncertainty
The Web Store listing establishes the extension’s advertised functions and setup, but not its reliability on complex sites, its handling of unusual login flows, or its success rate on multi-step transactions. It does not explain the permission model in technical detail, provide audit records, or specify how users approve consequential actions such as submitting a form or placing an order. The privacy disclosure identifies authentication information and website content as handled data, but the listing alone does not establish how that information is retained, isolated, or governed within an organization. Compatibility is stated for Chromium browsers; the source does not claim support for non-Chromium browsers.
Skill to build now
Create a small browser-agent access policy before connecting any production account. Use a dedicated browser profile, list the domains and account roles that the agent may access, and separate read-only research from actions that send, submit, purchase, or change records. Then test one bounded workflow, such as gathering invoices from a dashboard into a draft report, with a human reviewing the final output. This gives the team a concrete baseline for deciding where browser automation is safe enough to expand.
Reader takeaway
Perplexity Computer can now work through a user’s chosen Chromium browser, including already authenticated sites, via a desktop-app extension. That makes browser agents more immediately useful for real operations—and makes browser-session permissions and review controls a first-class part of deploying them.
Open question
How much visibility and approval will organizations get over actions taken through an authenticated local browser session, especially for workflows that modify records or transmit information?