Browser Automation

WinClaw has built-in browser automation capabilities using CDP (Chrome DevTools Protocol) to control your browser, letting the AI assistant help you with various web operations.

Supported Browsers

Currently supported browsers:

• Google Chrome
• Microsoft Edge

Other browsers are not supported yet. Please ensure you have Chrome or Edge installed.

Install agent-browser Tool

Before using browser automation, you need to install the agent-browser tool from the Tool Market:

1. Go to the Tools page and click Tool Market
2. Find agent-browser and click install

Once installed, you can start using the browser automation feature.

Start the Browser

Enter a command like this in the chat:

Start browser with agent-browser

The AI assistant will automatically complete the following steps:

1. Check existing browsers: Test CDP connection on port 9222 to confirm no browser instances are running
2. Start browser: Use agent-browser daemon start --browser chrome command to launch Chrome with remote debugging enabled
3. Verify status: Use agent-browser daemon status to confirm the browser daemon is running properly
4. Test functionality: Open a sample webpage to verify browser automation is working

After successful startup, you'll see a new browser window open — this is the browser that AI can control.

Manual Start (Optional)

You can also manually start the browser via the command line. Navigate to the tools installation directory and run the following commands:

1. Start the browser:

.\agent-browser.exe daemon start --browser chrome

2. Check the browser status:

.\agent-browser.exe daemon status

If the status shows no errors, the browser is successfully started and ready. You can go back to WinClaw and start using the browser automation feature.

Use the Browser

Once the browser is running, you can use natural language to have AI help with various tasks:

Open Xiaohongshu

The AI will automatically open the corresponding website:

More Examples

Maintain Login Status

The AI-controlled browser uses your real Chrome profile, which means:

• Cookies preserved: Logged-in websites stay logged in
• Sessions preserved: Browsing history, bookmarks are retained
• Settings preserved: Your browser personalization is not affected

Tip: Log in to Frequently Used Websites First

To help AI better complete automation tasks, we recommend manually logging into your frequently used websites after the first launch:

• Shopping sites like Amazon, eBay
• Social platforms like Twitter, Instagram
• Utility sites like email, cloud storage
• Company internal systems

Once logged in, AI can subsequently operate these websites in the logged-in state.

Keep the Browser Running

Recommendation: Don't close the AI-controlled browser window after starting.

Reasons:

• Closing it requires restarting agent-browser
• Keeping it running allows AI to execute browser tasks anytime
• Login status will remain active

If you temporarily don't need the browser, minimize the window instead of closing it.

Common Commands

AI uses these commands behind the scenes to control the browser:

You don't need to remember these commands — just describe what you want to do in natural language, and AI will automatically choose the appropriate command.

FAQ

Port occupied error when starting?

This means another browser instance is running. Close all Chrome/Edge windows and try again.

Browser started but AI can't control it?

1. Confirm the browser was started through AI, not opened manually
2. Check if CDP port 9222 is working properly
3. Re-enter the start command

What if a website requires login?

For first-time use, manually log in after AI opens the website. The login status will be saved, and AI can subsequently operate in the logged-in state.

How to use Edge instead of Chrome?

You can specify Edge:

Start Edge browser with agent-browser

Can I control multiple browser windows simultaneously?

Currently supports a single browser instance. For multiple pages, AI will open multiple tabs in the same browser.