Google previews Chrome auto browse, a Gemini agent that does multi-step web tasks, for US AI Pro and Ultra users

Google DeepMind/Gemini in Chrome auto browse

Google added an agent mode to Gemini in desktop Chrome: you describe a chore and it opens its own tab and scrolls, clicks and types to finish it. It launched as a preview for paying AI Pro and AI Ultra subscribers in the US, built on Gemini 3. Purchases and social posts need the user's confirmation.

• Runs in a separate tab marked with a cursor-and-sparkle badge; steps show in the side panel.
• Examples: shopping within a budget, booking appointments, filling forms, comparing travel options.
• Daily action limits apply; password autofill needs manual approval.
• Users can take over the task at any point.