OpenAI on October 15, 2026 extended computer use to its ChatGPT Chrome extension, which can now look across open tabs, take control of tabs to complete tasks, and organize tabs into groups. The Chrome Web Store lists the extension at 7,000,000 users, which gives the update immediate distribution. For business users, this turns the browser into a place where research and routine web work happen side by side.

OpenAI adds computer use to Chrome via ChatGPT extension

Chat answers, Work acts, tabs become context

In Chat mode, the assistant looks across open tabs to gather information and make answers more useful, according to the announcement on X. In Work, its agent for longer tasks, ChatGPT can control Chrome tabs for jobs such as sending messages, filling out forms, and downloading files. The company also said the tool can organize and group open tabs. The updated extension is available for all ChatGPT plans through the Chrome Web Store.

The store listing, version 1.26.1005.11759 updated October 15, 2026, describes background work without taking over the active session. It says ChatGPT on desktop and web can pull in context and take action in tabs, with access to installed plugins, full conversation history, and the local file system if connected. ChatGPT will always ask for permission when referencing browser history and when downloading or uploading files. Listed capabilities include research across live and authenticated sites, document and spreadsheet work, and data entry into timesheets, inventory, and content management systems.

Setup runs through the Computer Use section of the desktop app settings, with support for Chrome, Edge, Brave, Opera, and Vivaldi. Users install the extension, confirm the browser shows a Manage option, then select the browser with an @-mention in a Work or Codex chat. Side chat is available in Chrome, Edge, Brave, and Vivaldi but not in Opera. The documentation says ChatGPT can read or act on sites where the user is already signed in, naming LinkedIn, Salesforce, Gmail, and internal tools as examples.

What browser agents change for operations

Sample prompts published by OpenAI point to procurement, sales operations, and admin work. They include comparing three couches by dimensions, material, delivery, and returns, comparing pricing pages against requirements, and moving vendors into a sheet with pricing, features, and source links. Other examples cover comparing contractors and drafting an email, filling a registration form from an event brief while stopping before submission, and completing a dentist form while adding the appointment to calendars. For small firms, this compresses comparison shopping and form filling into one workflow.

Controls and data handling will shape adoption decisions. By default ChatGPT asks before interacting with each new website, with Allow once, Allow for this site, Allow for all sites, or Decline, plus shared allowlists and blocklists. Allow for all sites and browser-history access are labeled Elevated Risk, and history access is scoped to the request with no always-allow option. The listing states user data is not sold, not used for unrelated purposes, and not used for creditworthiness, while install prompts can include history, bookmarks, downloads, tab groups, and debugger access.

The marker to watch is whether tab-aware Work moves from demos to repeated use in authenticated systems. OpenAI retired its Atlas browser on August 9, 2026 and steered Chrome users to the extension, then expanded it to four more browsers on August 31. If companies standardize allowlists, connect plugins and files, and keep the agent on timesheets, expenses, and content systems, the browser becomes a managed work surface rather than a viewing window.