browser-automation

name	browser-automation
description	Use the OpenClaw browser tool for research, form-filling, and UI automation. Use when web_fetch is insufficient (JS-heavy pages, multi-tab research, authenticated flows).
metadata	{"nanobot":{"emoji":"🌐","os":["darwin","linux"]}}

Browser Automation

Use the browser tool for agentic web workflows. Default profile: openclaw (sandbox).

When to use

• Research across >3 tabs or JS-heavy pages
• Authenticated flows (login, form-fill)
• UI automation (click, type, screenshot, PDF export)

Prefer web_fetch for simple text from a single URL.

Common patterns

Screenshot page

1. browser(action="start") if needed
2. browser(action="snapshot", url="https://example.com") — navigate and snapshot
3. browser(action="screenshot") — capture visual

Click + type flows

1. browser(action="snapshot", url="...") — get element refs (e.g. e1, e2)
2. browser(action="act", kind="click", ref="e1") — click
3. browser(action="act", kind="type", ref="e2", text="...") — type into field
4. Re-snapshot after DOM changes — refs become stale

PDF export

1. Navigate to page
2. Snapshot to confirm content loaded
3. Use browser print-to-PDF if available, or screenshot as fallback

Workflow

• Snapshot first — generates refs before any act
• Re-snapshot after mutations — clicks, form fills invalidate refs
• Profile — use profile="openclaw" (default sandbox)

Examples

browser(action="snapshot", url="https://docs.openclaw.ai")
browser(action="act", kind="click", ref="aria-login")
browser(action="act", kind="type", ref="e3", text="username")

References

• docs/browser-automation.md — config and E2E test