Documentation

Start browser with CDP enabled. Uses an isolated profile so your personal browser is never touched.

Auto-detect browsers on the system and create an isolated profile directory.

Check if a CDP-enabled browser is running and reachable on the configured port.

Stop the browser instance managed by cdpilot on the current port.

Show the installed cdpilot version.

Set, show, or clear the HTTP proxy. Requires a browser restart to take effect.

Enable or disable headless mode. Requires a browser restart.

Navigate to a URL. Launches the browser automatically if it is not running.

Get the text content of the current page (document.body.innerText). Truncated at 1 MB.

Get the full HTML source of the current page (outerHTML). Truncated at 1 MB.

Take a full-page screenshot and save it as PNG.

Take a screenshot with green @N badges overlaid on every interactive element (buttons, links, inputs). Useful for AI agents to identify clickable targets.

Save the current page as a PDF file.

Click an element by CSS selector. Auto-waits up to 5 seconds for the element to appear, then scrolls it into view.

Fill an input field with a value. React/Vue compatible -- uses the native setter and dispatches input + change events.

Alias for fill. Type text into an input element (React/Vue compatible).

Submit a form by clicking its submit button or calling form.submit(). Defaults to the first <form> on the page.

Move the mouse cursor to the center of an element. Triggers CSS :hover styles and mouseover events.

Double-click an element by CSS selector.

Right-click (context menu) an element by CSS selector.

Drag an element from one position to another. Simulates mousedown, mousemove steps, and mouseup.

Send a keyboard shortcut or key press. Supports modifiers (ctrl, shift, alt, meta) combined with keys.

Scroll the specified element into view with smooth scrolling.

Wait for an element to appear in the DOM. Uses MutationObserver for efficient detection.

Click an element by its @N reference number from the last a11y-snapshot. Uses real mouse events via CDP Input domain.

Capture and display browser console logs (log, error, warning, info) from the current or specified page.

Monitor and display all network requests with status codes, types, and URLs.

Full auto-debug: navigate, then display console logs, network requests, performance metrics, and a screenshot -- all in one command.

Execute JavaScript in the browser and return the result. Supports async/await expressions.

Show key performance metrics: DOM nodes, JS heap size, event listeners, first meaningful paint, and DomContentLoaded.

Execute JavaScript across all open tabs in parallel and display each result.

List all open browser tabs with their index, title, and URL.

Open a new browser tab.

Close a tab by index or ID. Closes the active tab if no argument is provided.

Switch to a tab by its index number or CDP target ID.

Show current session info (session ID, target, status).

List all active sessions with their target IDs and status.

Close a specific session window and remove its registry entry.

Simulate network conditions: 3G, offline, or custom bandwidth/latency.

Intercept network requests: block URLs, mock responses from files, or inject custom headers.

List cookies for the current page or a specific domain.

Display all localStorage entries for the current page (keys and truncated values).

Emulate a mobile device (screen size, scale factor, user agent). Presets: iphone, ipad, android. Use 'reset' to go back to desktop.

Override geolocation. Presets: istanbul, london, newyork, paris, tokyo. Custom coordinates also supported.

Grant, deny, or reset browser permissions (geolocation, notifications, camera, microphone, etc.).

Start the MCP (Model Context Protocol) server over stdin/stdout. Enables AI agents like Claude Code and Cursor to control the browser via JSON-RPC.

Output a compact accessibility snapshot for AI agent navigation. Each line shows @ref [role] "name" with attributes. Use click-ref to interact.

Analyze the full accessibility tree. Subcommands: full (default), summary, find <role>.

Read a JSON array of commands from stdin and execute them sequentially. Returns results as JSON. Ideal for AI agent pipelines.

Toggle the visual indicator overlay. When enabled, shows a green border and cursor animations during automation.

Upload a file to a file input element using DOM.setFileInputFiles.

Handle JavaScript dialogs (alert, confirm, prompt). Auto-accept, auto-dismiss, or respond to prompts.

Access iframes and Shadow DOM. List iframes, evaluate JS inside them, or read shadow root content.

Configure the browser download directory or check its status.

List all installed browser extensions and dev mode extensions.

Install an extension from a CRX file or unpacked directory. Dev extensions are loaded on next browser launch.

Remove an extension by its ID or index number.

List all registered cdpilot project instances with their ports and status.

Stop a specific project's browser instance by name or path.

Stop all active cdpilot browser instances across all projects.

Close the currently active browser tab.