OpenAI Launches Chrome Extension for Codex, Enabling AI Agents to Navigate Live Web Sessions
OpenAI has released a new Chrome extension for its Codex AI coding agent, allowing it to operate directly inside a user's live browser session. The extension, announced Thursday, lets agents interact with signed-in websites, multiple tabs, and authenticated workflows without taking over the entire desktop.
This launch marks a shift from earlier “computer use” systems that relied on screenshot-and-click loops or plugin integrations. Instead, the extension connects Chrome to the Codex app on Windows and macOS, giving agents access to tools like Gmail, Salesforce, LinkedIn, and internal web apps using the user’s existing browser state and cookies.
Key Features
The extension enables Codex to work across multiple tabs and logged-in sessions in parallel. It avoids the traditional “screenshot, reason, move the mouse” cycle common in other computer-use systems, according to OpenAI developer experience lead Dominik Kundel.
“The new extension avoids the traditional screenshot, reason, move the mouse loop common in many computer-use systems,” Kundel said in a demo video.
Previously, Codex relied on structured plugins or broader computer-use tooling to interact with browser workflows. Plugins were preferred for services with APIs, but many workflows inside full web applications, internal dashboards, or authenticated sessions remained inaccessible.
Background
The launch builds on OpenAI’s “computer use” capabilities introduced in Codex in April. Those earlier capabilities allowed agents to operate desktop apps and browsers in the background while users worked elsewhere.
However, the new Chrome extension draws a clearer distinction between generalized computer-use systems and a more browser-focused approach. Instead of treating the browser as just another desktop app, Codex now integrates directly into Chrome’s internal processes.
AI companies have been racing to create coding agents that interact with software the way humans do: clicking buttons, scrolling pages, and moving cursors. While the promise is obvious, execution has often felt clunky, with agents monopolizing browser sessions or working through tasks one screen at a time.
What This Means
For enterprise users, the extension could significantly streamline automation of web-based workflows. Agents can now handle tasks across multiple authenticated services without requiring API integrations or manual login steps.
It also means that software developers and IT teams may be able to automate repetitive tasks inside internal tools that lack clean APIs. However, the extension is still new, and its reliability in complex, multi-step workflows remains to be seen.
OpenAI has not disclosed pricing or availability details beyond the initial launch. The extension is available now for Chrome users running the Codex app on Windows or macOS.