Ugot Browser Agent

Lets your AI agent work in your logged-in browser. Page content and screenshots are shared with the agent.

Ugot Browser Agent connects your own AI agent (Claude Code, Codex, Cursor and other MCP clients) to the Chrome you already use. The agent reads pages, clicks, types and captures screenshots with the sign-ins already in this Chrome profile. WHAT IT DOES - Works on any site out of the box. To keep the agent away from a site, limit it in Chrome's site access settings for this extension; limited sites and their frames are never read or controlled. - Task tab groups: tabs the agent opens are grouped by task name and color, marked ⌛ while working and ✅ when done. Your own tabs and groups are never moved. You can turn grouping off in the popup. - Visible control: a page border and virtual cursor show the agent at work. Stop it any time with the Stop button or Esc. - Approvals: in the default approval mode, submit, pay, delete and send-like clicks ask you first on the page. Typing into password fields is always refused. - Several Chrome profiles: install it in each profile you use; the agent can list them and open the one it needs. - Mail connection: from the popup, connect Gmail (Google sign-in) or Naver, Kakao and Daum mail (guided steps and an app password) to UGOT Mail. You type the app password in the popup; it is never given to the AI agent. HOW TO USE 1. Install the extension and the local bridge (ugot-browser-bridge 0.2.5 or later) as an MCP server in your AI client. 2. Open the popup and sign in to UGOT. 3. Ask your agent to work in your browser. DATA The extension talks only to the local bridge on 127.0.0.1. Page text, element information and requested screenshots of the pages the agent works on go through the bridge to the AI agent you connected, and may be processed by that AI provider. Mail settings you connect go to UGOT Mail on this computer and are stored encrypted in your UGOT account. Privacy policy: https://agent.ugot.uk/privacy.html