Using the Browser Automation Toolkit

OpenClaw comes with a powerful browser automation toolkit that turns your browser into a controllable tool for agents. Whether you're navigating sites, extracting data, or interacting with web apps, it's designed to be reliable, secure, and easy to use.

What the Toolkit Does

At its core, OpenClaw's browser automation lets agents perform actions just like a human would: opening tabs, clicking buttons, typing text, uploading files, and reading content. But unlike manual browsing, it's deterministic, scriptable, and runs in isolation.

Key capabilities:

Open and control tabs programmatically
Capture screenshots (full page or element-specific)
Take structured snapshots of UI elements for interaction
Click, type, select, drag, and submit forms
Wait for conditions (URLs, text, load states, or custom JavaScript)
Manage cookies, storage, geolocation, and viewport settings
Work with iframes and file uploads

It's built on Chromium's DevTools Protocol (CDP), with Playwright handling advanced actions, giving you both raw control and high-level ease.

Two Modes: Managed vs. Extension

OpenClaw supports two main ways to control the browser:

1. Managed Browser (Recommended for Automation)

The openclaw profile is a dedicated, isolated browser instance managed directly by OpenClaw. It doesn't interfere with your personal browsing and is ideal for automation tasks.

Runs in its own profile (orange accent by default)
Automatically launched and controlled
Perfect for headless or background workflows
Uses ports starting at 18800 to avoid conflicts

Start it with:

openclaw browser --browser-profile openclaw start

2. Chrome Extension Relay (Use Your Existing Browser)

The chrome profile lets you control your actual Chrome tabs through a browser extension. You manually attach it by clicking the OpenClaw icon on a tab.

No need to switch browsers
You decide when a tab is controlled
Great for manual debugging or live demos
Requires installing the OpenClaw Browser Relay extension

To set it up:

openclaw browser extension install

Then go to chrome://extensions, enable Developer mode, and "Load unpacked" using the path printed by the command above.

How to Use It

Automation works in three phases: navigate, inspect, and act.

Step 1: Launch and Navigate

First, ensure the browser is running:

openclaw browser start

Then open a page:

openclaw browser open https://example.com

Step 2: Inspect the Page

Take a snapshot to see clickable elements:

openclaw browser snapshot --interactive

This returns a numbered list of interactive elements. For more detail, add --labels to overlay reference numbers on a screenshot.

Step 3: Perform Actions

Use the reference numbers from the snapshot to click or type:

openclaw browser click 12
openclaw browser type 23 "Hello World" --submit

You can also wait for conditions before acting:

openclaw browser wait --text "Login successful"

Security and Isolation

Your personal browsing is never touched. The managed openclaw profile uses a separate user data directory. Even if automation logs into sites, it's sandboxed from your main browser.

Remote control (like via Browserless) is supported, but always keep CDP URLs and tokens secure. Treat them like passwords.

Debugging Tips

When something goes wrong:

Run snapshot --interactive to see current UI refs
Use highlight <ref> to see exactly what element will be targeted
Check console --level error for JavaScript issues
For advanced issues, record a Playwright trace with trace start and trace stop

Getting Started

Try this quick test:

openclaw browser start
openclaw browser open https://httpbin.org/forms/post
openclaw browser snapshot --interactive --labels
# Pick a field ref, then:
openclaw browser type <ref> "Test input"
openclaw browser click "Submit"

You'll see form automation in action. From there, scale up to login flows, data scraping, or end-to-end testing.

The browser toolkit turns the web into a first-class platform for agents. Use it wisely.