Skip to main content

Cloud Browser

With the cloud browser, the model can open pages that need scripts or clicks in a real browser running in the cloud. Use it for web scraping, data extraction and browser-driven work that a plain page read can't handle. This page shows how to make sure it's on and how to ask for it.

It's on by default​

The cloud browser comes as a tool and a Skill, and both are part of the default capabilities, so new chats start with them switched on:

  • The Cloud browser tool (under Find information): "Open pages that need scripts or clicks in a real browser."
  • The cloud-browser Skill: controls a cloud-hosted Chrome browser for web automation. It supports navigation, interaction, form filling, screenshots and multi-step web workflows. The model uses it when you ask for browser help, when browser automation suits the task better, or when other ways of fetching the page fail.

Check it's switched on​

If you changed your defaults or this chat's settings, check both:

  1. In the composer, click Tools and make sure Cloud browser is switched on.

    Tools panel with Cloud browser switched on

  2. Switch to the Skills tab and make sure cloud-browser is switched on.

    Skills panel with cloud-browser switched on

Changes in these panels apply only to the current chat. See Tool calling and Use Skills.

Ask the model to use it​

Say in your message that you want the cloud browser, for example:

Use the cloud browser to visit https://www.example.com and extract the page title and all links.

The model loads the cloud-browser Skill and works through the page step by step. The tool calls and results appear in the chat.

For pages that load fine without scripts, the Read web pages tool ("Open a URL and extract its main text") is usually enough.

Use it in tasks and agents​

Tasks and agents keep their own capabilities. To give one the cloud browser, add the Cloud browser tool and the cloud-browser Skill with Add capabilities, or click Use my default capabilities.