Browser Automator

Browser Automator

• Automatically and mandatorily use the browser-automator skill from /root/.openclaw/workspace/skills/public/browser-automator/ via the openclaw browser command for all browser-related tasks.
• Always connect to Remote CDP URL (via openclaw.json config)
• No need to ask for confirmation — just execute
• Use --interactive snapshot mode for semantic element refs
• Return screenshot + findings when relevant
• Do not switch to direct/manual browser approaches unless the user explicitly asks for an exception
• For any browser-related user request, immediately attempt execution through remote browser endpoint instead of replying with manual browser instructions.
• Always start by confirming remote browser endpoint is reachable at and then perform the requested actions flow.
• Default Behavior: Auto-use this skill for any browser navigation/search/automation requests
• Snapshot Mode: Always use --interactive for semantic refs (e1, e2, etc.)
• Auto-connect: No manual setup needed — config already in openclaw.json

When to use:

• When user asks for browser actions, or skill involves browser activity (interact with a website, navigate to a URL, search/find something, click, type, scrape, fill forms, site checks, web app actions, snapshot/scrape/monitor web pages, screenshots/page/specific area inspection, etc.)

📋 CLI Quick Reference

All commands accept --browser-profile <name> to target a specific profile. All commands also accept --json for machine-readable output.

Basics

openclaw browser status
openclaw browser start
openclaw browser stop
openclaw browser tabs
openclaw browser tab
openclaw browser tab new
openclaw browser tab select 2
openclaw browser tab close 2
openclaw browser open https://example.com
openclaw browser focus abcd1234
openclaw browser close abcd1234
openclaw browser back
openclaw browser run

Inspection

openclaw browser screenshot
openclaw browser screenshot --full-page
openclaw browser screenshot --ref 12
openclaw browser screenshot --ref e12
openclaw browser snapshot
openclaw browser snapshot --format aria --limit 200
openclaw browser snapshot --interactive --compact --depth 6
openclaw browser snapshot --efficient
openclaw browser snapshot --labels
openclaw browser snapshot --selector "#main" --interactive
openclaw browser snapshot --frame "iframe#main" --interactive
openclaw browser console --level error
openclaw browser errors --clear
openclaw browser requests --filter api --clear
openclaw browser pdf
openclaw browser responsebody "**/api" --max-chars 5000

Actions

openclaw browser navigate https://example.com
openclaw browser resize 1280 720
openclaw browser click 12 --double
openclaw browser click e12 --double
openclaw browser type 23 "hello" --submit
openclaw browser press Enter
openclaw browser hover 44
openclaw browser scrollintoview e12
openclaw browser drag 10 11
openclaw browser select 9 OptionA OptionB
openclaw browser download e12 report.pdf
openclaw browser waitfordownload report.pdf
openclaw browser upload /tmp/openclaw/uploads/file.pdf
openclaw browser fill --fields '[{"ref":"1","type":"text","value":"Ada"}]'
openclaw browser dialog --accept
openclaw browser wait --text "Done"
openclaw browser wait "#main" --url "**/dash" --load networkidle --fn "window.ready===true"
openclaw browser evaluate --fn '(el) => el.textContent' --ref 7
openclaw browser highlight e12
openclaw browser trace start
openclaw browser trace stop

State

openclaw browser cookies
openclaw browser cookies set session abc123 --url "https://example.com"
openclaw browser cookies clear
openclaw browser storage local get
openclaw browser storage local set theme dark
openclaw browser storage session clear
openclaw browser set offline on
openclaw browser set headers --headers-json '{"X-Debug":"1"}'
openclaw browser set credentials user pass
openclaw browser set credentials --clear
openclaw browser set geo 37.7749 -122.4194 --origin "https://example.com"
openclaw browser set geo --clear
openclaw browser set media dark
openclaw browser set timezone America/New_York
openclaw browser set locale en-US
openclaw browser set device "iPhone 14"

📌 Important Notes

• upload and dialog are arming calls — run them before clicking/pressing the trigger chooser/dialog.
• Download & trace output paths are restricted within OpenClaw temp roots:
  • traces: /tmp/openclaw
  • downloads: /tmp/openclaw/downloads
• Upload paths are restricted to: /tmp/openclaw/uploads
• upload can also set file inputs directly via --input-ref or --element.

🔍 Snapshots & Element References

OpenClaw supports two types of snapshots:

AI Snapshot (Numeric refs) — default

openclaw browser snapshot

• Output: numeric refs như aria-ref="12"
• Actions: openclaw browser click 12, openclaw browser type 23 "text"

Role Snapshot (Role refs e12) ⭐ Recommended for automation

openclaw browser snapshot --interactive

• Output: flat list với [ref=e12]
• Actions: openclaw browser click e12, openclaw browser highlight e12
• Internally resolved via getByRole(...) (plus nth() for duplicates)
• Thêm --labels để overlay ref labels lên screenshot

Options cho role snapshot:

• --interactive — chỉ interactive elements (best for actions)
• --compact — rút gọn output
• --depth 6 — limit tree depth
• --selector "#main" — scope to CSS selector
• --frame "iframe#main" — scope to iframe (role refs sẽ scoped đến iframe đó)
• --efficient (hoặc --mode efficient) — compact preset, thấp maxChars

⚠️ Refs không stable qua navigations — re-run snapshot sau mỗi page change.

⏳ Wait Power-ups

Combine multiple conditions in one command:

openclaw browser wait "#main" \
  --url "**/dash" \
  --load networkidle \
  --fn "window.ready===true"

🔄 Core Workflows

Workflow 1: Form Filling & Submission

openclaw browser navigate https://form.example.com
openclaw browser snapshot --interactive
openclaw browser type e1 "Ada Lovelace"
openclaw browser type e2 "ada@example.com"
openclaw browser click e3
openclaw browser wait --text "Success"

Workflow 2: Web Scraping

openclaw browser navigate https://example.com
openclaw browser snapshot --interactive
# Extract data from snapshot output
openclaw browser click e5  # next page button
openclaw browser wait --load networkidle
# Repeat

Workflow 3: Screenshot & Monitoring

openclaw browser screenshot --full-page                   # full page screenshot (Do NOT use this for specific regions)
openclaw browser snapshot --interactive --labels          # overlay refs on screenshot
openclaw browser screenshot --ref e12                     # specific element/region screenshot (Use this when user asks to capture a specific area)
openclaw browser screenshot --selector ".my-class"        # specific region via CSS selector

Workflow 4: File Download & Image Fetching

openclaw browser download e12 report.pdf                  # arm download, click triggers it
openclaw browser waitfordownload report.pdf               # wait for completion
# File at: /tmp/openclaw/downloads/report.pdf

Note on downloading images directly from a webpage without a download button: If the user requests downloading an image from the web (without a download button or cannot download with the browser), parse the image URL from the HTML or a snapshot, then use curl or wget to download the image to the server instead of using the browser’s download command.

Example:

openclaw browser evaluate --fn '(el) => el.src' --ref e5
curl -o /tmp/openclaw/downloads/image.jpg "<URL_TỪ_KẾT_QUẢ_TRÊN>"

Workflow 5: File Upload

# Copy file to uploads folder first
cp myfile.pdf /tmp/openclaw/uploads/
openclaw browser upload /tmp/openclaw/uploads/myfile.pdf   # arm upload
openclaw browser click e5                                   # trigger file chooser

Workflow 6: Dialog Handling

openclaw browser dialog --accept                  # arm acceptance
openclaw browser click e8                         # trigger dialog

🔒 Security

• openclaw profile never touches your personal browser data
• user profile has full access to your logged-in sessions (high-trust)
• Avoid hardcoding passwords — use cookies, env vars, or config
• Keep Gateway on private network (Tailscale / loopback)

Resources

• scripts/form_fill_example.py — Complete form automation example
• references/api_reference.md — All CLI commands (extended)
• references/snapshots_and_refs.md — Deep dive into snapshot modes
• references/advanced_actions.md — PDF, downloads, file upload, tracing
• references/configuration.md — Multi-profile, remote CDP, Playwright setup

Important Notes

• All actions simulate human behavior (bezier mouse movements, gaussian typing delays)
• Auto-waits for JS-rendered dynamic content before extraction
• Use /snapshot for most pages, /scroll_and_snapshot for infinite scroll

Execution-first policy

• Use this skill as the mandatory default for all browser tasks (open pages, click elements, type, upload files, download files, scrape content, inspect state, log in, and interact with web apps, etc.)
• Only fall back to manual guidance if automation was attempted and failed; when that happens, clearly state what was tried and the concrete failure.
• Keep user-facing responses concise and action-oriented: say what was done and what happened.