Agents

OpenAI ChatGPT Chrome Tool Can Now Read and Act On Tabs

OpenAI has upgraded its ChatGPT Chrome extension, allowing the assistant to directly interact with, analyze, and control open browser tabs across all user subscription plans.

The Decoder16 hrs agoAgents
Image: The Decoder

OpenAI has expanded its ChatGPT extension for Google Chrome into a full-fledged browser assistant capable of reading, analyzing, and directly manipulating open tabs. Users can call upon the assistant using the @chrome prompt handle to target specific active browser windows. The updated extension introduces two main operational features: Chat mode, which extracts and synthesizes data across multiple open tabs simultaneously, and Work mode, which leverages autonomous control mechanics to perform actions within the browser environment.

Through Work mode, the assistant uses Computer Use technology to execute complex digital workflows. It can automatically fill out online forms, dispatch web messages, trigger file downloads, and organize browser real estate by grouping open tabs. Powering these action-oriented capabilities is GPT-6 Astra, which currently stands as the leading Computer Use model on the market for completing detailed, multi-step tasks. Additionally, the interface has been redesigned with an upgraded sidebar that displays a user's full chat history side by side with their current web page.

Practical examples provided for the tool illustrate its utility in everyday research and e-commerce tasks. The extension can compare items across distinct open storefront tabs by evaluating specific attributes such as product size, material composition, estimated shipping times, and return policies. It can also condense long-form online articles into plain prose and normalize pricing pages across different commercial vendors to streamline price comparison.

The upgraded extension is currently available to users across all ChatGPT plans. For developers, researchers, and general users, this integration marks a significant shift away from manually copying and pasting context back and forth between web applications and a separate AI chat window. By placing an autonomous model directly into the web client, OpenAI enables more streamlined web research and automated administration straight from the browser tab bar.

This is our own summary of reporting by The Decoder

More in Agents