Browser Subagent

Fresh

The browser subagent is invoked when the agent needs to interact with the browser. It uses a specialized model (different from the main agent's reasoning model) optimized for visual understanding and browser control.

Capabilities

The browser subagent has access to the following tools:

Tool	Description
Clicking	Click on page elements identified visually or via DOM
Scrolling	Scroll pages to reveal content
Typing	Enter text into form fields and inputs
Console Logs	Read browser console output
DOM Capture	Inspect and extract DOM structure
Screenshots	Capture page or element screenshots
Markdown Parsing	Extract and parse page content as markdown
Video Recording	Record agent actions as playable recordings

Visual Indicators

• A blue border overlay appears on the browser window when the subagent is active
• The overlay indicates the subagent is controlling the browser -- do not interact manually during this time
• The subagent can act on unfocused tabs -- it does not require the tab to be in the foreground

TIP

The browser subagent produces screenshots and recordings as artifacts. Review these in the Agent Manager to verify the subagent performed the expected actions.