Open Browser Control

Give AI agents control of your Chrome browser. Works with Claude, Codex, Cursor, and any MCP client.

Description:

Give AI agents control of your real Chrome browser — your cookies, sessions, and logins. When the AI hits something it can't handle (sign-in, CAPTCHA, MFA), it asks you to step in, then continues where it left off.

Works with Claude Code, OpenAI Codex, Claude Desktop, Cursor, and any MCP-compatible client.

HOW IT WORKS

Your AI agent connects to Chrome through a lightweight bridge. The agent gets 19 browser tools — navigation, clicking, typing, DOM inspection, JavaScript execution, tab management, and more. Everything runs locally through Chrome DevTools Protocol. No data leaves your machine.

SETUP

1. Install this extension
2. Add the MCP server to your AI client: npx -y open-browser-control
3. Start your agent — it connects automatically

Full setup instructions: https://github.com/smankoo/open-browser-control

KEY FEATURES

- Uses your real browser with your existing sessions and logins
- User/AI handoff: AI pauses and asks you to take over when needed
- Per-task tab groups: each task gets its own named Chrome tab group
- DOM-first browsing: reads page structure directly instead of relying
on screenshots
- Side panel UI for monitoring sessions and switching control modes
- Works with any MCP client — not locked to a single AI provider

TOOLS AVAILABLE TO THE AI

Navigation, DOM inspection, JavaScript execution, clicking, typing, scrolling, keyboard input, hovering, dropdown selection, waiting, screenshots, tab management (open, close, switch, list), tab group creation, user handoff for authentication flows.

OPEN SOURCE

MIT licensed. Source code:
https://github.com/smankoo/open-browser-control