A firewall for browser agents
AI Browser Guard is a Chrome extension that detects, monitors, and controls AI agents operating in your browser. It sees the agent without the agent identifying itself, blocks actions outside the boundary, and ends every connection with a kill switch. Open source under Apache-2.0.
Why a firewall
A firewall watches traffic, applies a policy, and stops what falls outside it. AI Browser Guard does that for agents in the tab. When a browser agent takes over a session, the extension detects it, checks each action against a delegation rule, and blocks the ones outside the boundary before they execute. The kill switch is the hard cutoff when a session needs to end now.
Detect
Three independent layers find an agent without it identifying itself: CDP debugger monitoring, V8 stack trace analysis, and environment fingerprinting. Catches Playwright, Puppeteer, Selenium, Anthropic Computer Use, and OpenAI Operator.
Block
Delegation rules define what an agent may do. An action outside the boundary is blocked before it runs, with an inline toast that names the action and offers allow once or whitelist this domain.
Kill switch
One click revokes permissions, clears automation flags, and terminates every CDP session. Keyboard shortcut Ctrl+Shift+K. Every action is logged to a session timeline for review.
An agent takes over the tab. AI Browser Guard detects it, blocks the transfer it tries to submit, and the kill switch ends every connection.
Install it
Add the extension from the Chrome Web Store, then pin it. Detection starts on the next page load, with no configuration required.
Open the Chrome Web Store listingPress Ctrl+Shift+K at any time to end every agent connection on the current tab.