On March 14, developer Petr Baudis stated that with the official release of Google Chrome version 146, users can now expose their current live browsing session to the AI Agent via the Model Context Protocol (MCP) through a switch, enabling the AI Agent to perform operations directly within the user's logged-in browsing environment.
Baudis also revealed that he had previously envisioned using the command `amp mcp add chrome-devtools --npx chrome-devtools-mcp@latest --autoConnect` to allow Claude to browse and perform tasks on behalf of the user while they were logged in. Now, with the implementation of Chrome MCP capabilities, this scenario is gradually becoming a reality.





