browser

Browser Automation

You MUST use the mcp__plugin_browser-for-ai_browser__* tools (NOT computer-use, NOT Claude in Chrome).

Instructions

Given the user's task description or URL:

1. Navigate using browser_navigate to go to the target URL
2. Take a screenshot using browser_take_screenshot to show the user what's on screen
3. Interact using browser_click, browser_type, browser_press_key, etc.
4. Verify with screenshots after important actions
5. Report back to the user with what was found or accomplished

IMPORTANT: Tool Selection

• ALWAYS use tools starting with mcp__plugin_browser-for-ai_browser__
• NEVER fall back to computer-use or Claude in Chrome tools
• If a browser tool fails, retry or report the error — do NOT switch to other tools

Available Tools

Navigation

• browser_navigate — Go to a URL
• browser_navigate_back — Go back

Interaction

• browser_click — Click an element by text or CSS selector
• browser_type — Type text into a focused element
• browser_select_option — Select from a dropdown
• browser_hover — Hover over an element
• browser_press_key — Press keyboard keys
• browser_fill_form — Fill form fields
• browser_mouse_click_xy — Click at specific x,y coordinates

Observation

• browser_take_screenshot — Capture the page as an image (SHOW THIS TO THE USER)
• browser_snapshot — Get accessibility tree (text-based, good for finding elements)
• browser_console_messages — Read browser console
• browser_network_requests — See network activity

Tabs

• browser_tabs — List open tabs

Best Practices

1. Always take a screenshot after navigating or performing important actions — this shows the user what's happening
2. Use browser_snapshot to understand page structure before clicking
3. Use text-based selectors for clicking — more robust than CSS selectors
4. Take screenshots frequently so the user can follow along