Use Browser Automation
Browser automation gives supported agents a browser tool for inspecting and interacting with web applications. It is disabled by default and should be enabled only for tasks and destinations you intend to expose to the agent.
Enable the browser tool
- Open AI Cockpit Reasoning: Settings.
- Select Web Tools.
- Turn on Enable Browser Automation.
- Keep Use System Chrome on to use the Google Chrome installed on your computer, or turn it off when a compatible Playwright Chromium is already installed.
- Turn on Headless Mode only when you do not need to watch the browser.
The same tab has Web Search, which makes web search available to models from every provider.
The extension makes the browser tool available after the configuration is saved. Start a new request if an existing agent does not detect the change.
Request a browser task
Provide the URL, expected state, and a stopping condition. For example:
Open http://localhost:3000, verify that the documentation navigation loads,
and report browser console errors. Do not submit forms or change data.
Review navigation, downloads, authentication, and data-changing actions through the normal permission flow.
Visible or headless execution
Use a visible browser when you need to observe interactions or complete authentication. Use headless execution for repeatable checks that do not require human input.
The browser can access data available to its active profile. Do not use a profile that contains unrelated personal sessions or customer data.