Give your AI Agent the ability to use your browser using browser-use

If you'd like AI Agent (Amazon’s Kiro in my case) to be able to use your browser, (navigate pages, click elements, take screenshots etc.), you can do so using the browser-use MCP server.

To install browser-use, follow the instructions below.

First, ensure that you’ve Python and uvx installed (Python is typically available on most operating systems).

uvx --version

If you don’t have Python or uvx installed, install both.

Next, install browser-use

uvx --from 'browser-use[cli]' browser-use --help

Then, add browse-use MCP server to your AI Agent.

{
  "mcpServers": {
    "browser-use": {
      "command": "uvx",
      "args": ["--from", "browser-use[cli]", "browser-use", "--mcp"]
    }
  }
}

If you use Kiro, you can add browse-use MCP server to the mcpServers property in ~/.kiro/settings/mcp.json or to the mcpServers property in your agent’s JSON in ~/.kiro/agents/. You can also optionally add @browser-use/*"to your Kiro agent JSON’s allowedTools array property to auto-approve all browser-use MCP server’s tools (so Kiro doesn’t prompt you approval when it’d like to any of those tools).

If you use Claude Code, add browser-use MCP server to ~/.claude/settings.mcp.json

Start your AI agent and ask it to browse a page. You should see a Chrome window open and Kiro navigating it.

Using Your Chrome Profile

By default, the MCP server uses an isolated browser profile. To use your real Chrome browser with all your saved cookies and logins, create a config file at ~/.browser-use/config.json:

{
  "profile": {
    "user_data_dir": "~/Library/Application Support/Google/Chrome",
    "profile_directory": "Default"
  }
}

The user_data_dir path varies by platform:

• macOS: ~/Library/Application Support/Google/Chrome
• Windows: %LOCALAPPDATA%\Google\Chrome\User Data
• Linux: ~/.config/google-chrome

Close Chrome completely before starting the MCP server, as Chrome locks its profile directory when running.