Chrome DevTools MCP adds a switch to block JS execution

Chrome DevTools MCP shipped version 1.9.0 on September 8, and the headline change is a set of switches to turn off JavaScript execution. The release adds an option to disable the JS execution tools outright and extends the existing --no-javascript-evaluation flag to cover navigations and initScripts, not just direct evaluation calls.

Until now, giving an agent browser access meant giving it arbitrary in-page code execution too. There was no way to separate the two. If you wanted an agent to click through a checkout flow or read a dashboard, it also had the ability to run whatever JavaScript it wanted on any page it opened.

That's a real gap if you're running agents against production sites or anything with session cookies and auth tokens in scope. A browser automation tool with unrestricted eval is a bigger attack surface than most teams account for when they wire it into a CI job or a long-running agent loop.

The same release adds --allow-unrestricted-paths and configurable filesystem roots, and raises the default trace buffer to 1.2GB to match DevTools itself. It also packages the tool as an Agent Plugins 1.0 release, with a version-pinned mcp.json tested through add, doctor, repair and remove passes against Claude Code, Gemini CLI, OpenCode, Cline and Windsurf.

The release notes don't say whether the JS-disabled mode still allows form fills and clicks, the difference between a browser agent that's merely constrained and one that can't do its job. Check that before flipping the switch on anything that needs to click or type rather than just read a page.