chrome-debug-mcp

Capture a screenshot of the current page. Returns a base64 encoded PNG image.

Clicks on an element in the DOM using a CSS selector

Fills an input field in the DOM with the specified text. It focuses the element using a CSS selector and then simulates native typing.

Evaluate JavaScript in the current Chrome tab

Navigate the current Chrome tab to a URL

Fetch the entire HTML of the current document or a snippet around a search query

Enable debugger and pause on the first statement of the next executed script, then reload the page

Step over the current line of code in the debugger

Resume execution in the debugger

Search all parsed scripts for a specific text/query and get the line and column number for setting breakpoints

Set a debugger breakpoint at a specific script, line and column

Evaluate an expression in the context of the currently paused debugger call frame

Reload the current page

Scrolls the page by a specific amount (pixels or pages) or to a specific element.

Remove a debugger breakpoint

Restarts the managed Chrome instance with remote debugging enabled. This tool can be used to start a new instance or restart an already existing one. If Chrome fails to start due to a missing executable, advise the user to set the CHROME_PATH environment variable for the MCP server.

Stops the managed Chrome instance. This tool should be used to ensure no zombie Chrome instances are left running after finishing operations with the MCP server.

Retrieve intercepted network requests (REST) and WebSocket frames with optional advanced filtering. Use this tool to inspect API calls, page assets, and WebSocket traffic. It is highly recommended to use the available filters to avoid flooding the context window with unnecessary network data.

Retrieve console logs from the browser. This includes console.log/warn/error calls, exceptions, and network errors. Use this tool for troubleshooting page scripts and errors.

Retrieve run-time performance metrics from the browser (e.g., JS heap size, DOM nodes, layout duration, etc.). Useful for getting a quick snapshot of the page's memory and computational overhead.

Record and analyze a performance trace of the page. It automatically calculates Core Web Vitals (FCP, LCP, DCL, Load) and identifies the top Long Tasks (main thread blocking operations). You can optionally reload the page with cache disabled to simulate a cold start.

Enables proxy authentication using the Fetch domain. Call this after restarting Chrome with a proxy-server if your proxy requires authentication. It maintains active listening for up to 30 seconds of inactivity before automatically unhooking itself.

EXPERIMENTAL: Send a raw CDP command to the browser. Use ONLY if existing specialized tools (like navigate, click_element, etc.) do not satisfy your needs. Requires knowledge of the Chrome DevTools Protocol (CDP).

EXPERIMENTAL: Retrieve events captured from the browser that are not handled by other specialized listeners (like network or console). Use this to see results of custom commands that produce events (e.g., Target.targetCreated).