Deeporax Browser MCP

Let AI agents see and control this browser: DOM snapshots, screenshots, clicks, typing, navigation, console and network.

Deeporax Browser MCP connects your own AI coding agent to the browser you already use, with the sessions you are already signed into. Install the companion MCP server, point your agent at it, and ask for what you want in plain language. The agent reads the page structure, clicks the right button, fills the form, and takes a screenshot so you can check the result. WHAT YOUR AGENT CAN DO - Read a page as a structured outline of its links, buttons, headings and inputs, so it acts on the right element instead of guessing - Click, type, hover, drag, select dropdown options, and press keys - Fill multi-field forms in one step - Take screenshots of the visible tab - Open, close, list and switch tabs, and navigate back and forward - Read console messages and network requests when debugging a page - Run JavaScript in the page and read the result - Wait for text or an element to appear before continuing YOU CAN SEE AND STOP IT While the agent is working, the tab shows an orange border, a status pill naming the current action, and a cursor that moves to each target before it clicks. A Stop button halts control immediately. WHERE YOUR DATA GOES Nowhere. The extension talks to a server on your own machine over 127.0.0.1 and nothing else. No accounts, no telemetry, no analytics, no remote code. Page content never leaves your computer. SETUP 1. Install the server: npm i -g deeporax-browser-mcp 2. Register it with your MCP client, for example: claude mcp add deeporax-browser-mcp -- deeporax-browser-mcp 3. Keep Chrome open. The toolbar badge reads ON when connected. Works with Claude Code, Claude Desktop, and any client that speaks the Model Context Protocol. Requires Node.js 18 or newer. Source code and documentation: https://github.com/imfaisii/deeporax-browser-mcp