The competition among AI companies, often referred to as the LLM war, intensifies to productivity.
What began in earnest after the public release of OpenAI's ChatGPT in late 2022, a contest developed ever more capable language models. But now, things have shifted toward building agents that can handle practical, multi-step work across the tools people already use every day.
OpenAI has now expanded one of its own agents with deeper browser integration.
The company announced that its Codex tool can now operate directly inside Chrome on macOS and Windows through a new extension.
Users install the plugin from within the Codex desktop app.
Once connected, Codex interacts with websites and applications in the browser without seizing control of the window or disrupting other tabs. It can run tasks across multiple tabs in parallel and in the background while the user continues working elsewhere.
Codex now works directly in Chrome on macOS and Windows.
It’s even better at working with apps and sites in Chrome, and now works in parallel across tabs in the background without taking over your browser.
To get started, install the Chrome plugin in the Codex app. pic.twitter.com/pjtHd9gC69— OpenAI (@OpenAI) May 7, 2026
Codex originated in 2021 as OpenAI’s specialized model for generating and understanding code. It first gained attention for powering early versions of GitHub Copilot and similar developer tools.
That early Codex focused primarily on translating natural language into code snippets, and later, as an autonomous, cloud-based software engineering agent. It was later phased out in favor of more general-purpose models, but the name has now been revived for a far more ambitious agentic system.
In other words, the name has now come to describe a broader AI assistant that plans, writes code, and executes tasks on a user's behalf.
And this Chrome extension builds on that foundation by letting Codex treat the browser as a native environment rather than an external interface.
With the new Chrome extension, Codex can quickly move through repetitive browser work, like navigating structured pages and complex data entry flows.
Under the hood, it writes and runs code to navigate and complete tasks. pic.twitter.com/6bfDlnK2U3— OpenAI (@OpenAI) May 7, 2026
In practice, the update allows Codex to manage repetitive browser-based work more fluidly.
For example, it can navigate structured pages, fill out forms, extract data, and complete flows that previously required manual switching between tabs or applications. It can also scan a community forum, pull relevant threads, summarize key points, and assemble the results into a spreadsheet. It can also check email inboxes, locate attached files, verify details against local documents, and upload everything into an expense-tracking system.
When a task requires coordination across several sites or tools, Codex selects the most suitable method for each step, sometimes using its existing plugins, sometimes the browser extension, and sometimes a combination of both.
This move reflects the broader direction the industry has taken.
Companies are no longer competing solely on model size or benchmark scores. The focus has moved to reliability in everyday environments, seamless handoffs between different software, and the ability to complete end-to-end workflows without constant human supervision.
Codex's Chrome integration is one concrete step in that direction, showing how an AI agent can quietly handle the kinds of digital chores that once consumed hours of clicking, copying, and context switching.
If a task needs multiple tools, Codex chooses the best one for each step.
It uses plugins when they can handle the job, Chrome when it needs a logged-in website, and combines approaches as needed. pic.twitter.com/3GvDouoPDi— OpenAI (@OpenAI) May 7, 2026
It's worth noting that the extension means that Codex can now operate directly inside the browser, where the work already happens.
With access to tabs, logged-in tools, and the ability to run tasks in parallel in the background, it feels far closer to a true operator than just a coding demo.
In other words, the breakthrough isn't simply Codex in Chrome, but Codex embedded within the tools and accounts users are already signed in into.
That means less context switching, faster feedback loops, and a much more practical fit for real-world workflows.
The Chrome extension expands what Codex can do for coding and work.
From debugging browser flows to checking dashboards, conducting research, or updating CRMs, Codex can take on more of the tasks that already happen in your browser.
Available today in the Codex app in all…— OpenAI (@OpenAI) May 7, 2026
It's also worth noting that the OpenAI's Codex Chrome extension is not the only effort in this space.
The company has previously introduced Atlas, a dedicated Chromium-based browser with ChatGPT built directly into it. While Codex extends agent capabilities into the user's existing Chrome browser, running quietly in background tabs and coordinating with other tools, Atlas offers a more unified, purpose-built environment.
It features a persistent sidebar for context, browser memory that retains details from visited sites, and an Agent Mode that handles navigation, clicking, and multi-step tasks within the dedicated application itself.
Together, these approaches reflect OpenAI's broader push toward practical agentic systems.
Codex emphasizes flexibility and minimal disruption for users who prefer to stay within familiar browsers, particularly for technical and productivity workflows. Atlas, by contrast, creates a more integrated AI-native browsing experience. In the near term, users may benefit from both, or see them gradually converge as OpenAI works toward unified desktop tools.
Ultimately, the LLM war that began with ChatGPT has evolved into a practical contest over how seamlessly AI can operate inside the digital environments people already use every day.
Further reading: With Computer Use On Mac, OpenAI Revives Codex As An 'Almost Everything' AI Agent