chrome-devtools-cli

A CLI-driven way to automate Chrome and use Chrome DevTools from the terminal or shell scripts, provided by chrome-devtools-mcp. The background server starts implicitly on first use, so the workflow is simply execute, inspect, act: run a tool such as chrome-devtools list_pages, capture element UIDs with take_snapshot, then act on them with commands like click, fill, hover, drag, type_text, or upload_file. State persists across commands, any command accepts --help, and output defaults to Markdown with --output-format=json available for structured results.

Use it when you want to script browser tasks rather than drive them interactively. The command surface is broad: navigation (navigate_page, new_page, select_page, list_pages, close_page), emulation of network conditions, CPU throttling, geolocation, color scheme, viewport, and user agent plus resize_page, and performance tooling such as performance_start_trace, performance_stop_trace, performance_analyze_insight, and take_memory_snapshot.

For inspection and debugging it exposes evaluate_script (including passing snapshot UIDs as arguments), console message listing and filtering, take_screenshot for the viewport, full page, or a single element, lighthouse_audit, and network request listing and retrieval with pagination and resource-type filters. It can also manage Chrome extensions by listing, installing from a path, uninstalling, reloading, and triggering an extension default action. Experimental tools such as click_at coordinate clicking and screencast recording are disabled by default and must be enabled with their flags during start. Installation is a one-time prerequisite documented separately and is not part of the regular workflow.