browsing-with-robot

Browser Automation with robot

Quick start

# start headless browser
robot start
# navigate to a page
robot navigate "https://example.com"
# extract page text (prefer over screenshot — saves tokens)
robot text
# interact with elements using CSS selectors
robot click "#my-button"
robot type "#email" "user@test.com"
# screenshot only when you need to see visuals
robot screenshot
# always clean up when done
robot stop

Commands

Browser lifecycle

robot start                    # headless default
robot start --headless=false   # visible browser
robot stop                     # always stop when done
robot status                   # current URL, title, state

Navigation & extraction

robot navigate <url>           # navigate and wait for load
robot text                     # extract visible page text
robot screenshot               # save PNG, returns file path

Interaction

robot click <selector>         # CSS selector (e.g. #btn, .link, button)
robot type <selector> <text>   # type into input field

Output

All responses are JSON. View screenshots with the Read tool.

{"ok": true, "url": "...", "title": "..."}
{"ok": true, "text": "..."}
{"ok": true, "path": "/tmp/robot/screenshot-1234.png"}
{"ok": false, "error": "...", "suggestion": "..."}

Example: Scrape page content

robot start
robot navigate "https://example.com"
robot text
robot stop

Example: Fill and submit form

robot start
robot navigate "https://example.com/form"
robot type "#email" "user@test.com"
robot type "#password" "secret"
robot click "button[type='submit']"
robot text
robot stop

Tips

• Start first. "daemon already running" error? Run robot stop then robot start.
• Prefer text over screenshot — text is token-cheap. Screenshot only for visual layout.
• robot stop runs automatically when the session ends, but you can also run it manually between tasks.