Your own Chrome, driven by a terminal agent.

What it does

• Your real profile

  Chromium 136 and later refuse --remote-debugging-port on your default profile, so the DevTools protocol cannot reach the browser you actually use. An extension runs inside that profile with your sessions instead.

• Its own tab group

  The agent only touches tabs it opened or adopted, kept in a group you can see. Close the group and the whole session is gone.

• Numbered snapshots

  chromectl snapshot turns the page into numbered elements. Click by number, type into a field, scroll, wait for a selector, run JS, take a screenshot.

• Nothing on the network

  The daemon binds 127.0.0.1 and the extension connects to it over a websocket. A ping every 20 seconds keeps the MV3 worker alive; nothing is exposed to your LAN.

• Any agent, one command

  The agent just runs chromectl ... in a shell, so no vendor's chat sidebar sits in the middle and any harness can drive it.

• Background by default

  Opening a page does not steal focus. Tabs land in the Agent group while you keep working in your own.

Install

1. Run the installer

   It installs dependencies, compiles the TypeScript, links chromectl into ~/.local/bin, links the skill for Claude Code and opencode, and starts the daemon.

   git clone https://github.com/ghostyshell/browser-bridge.git
   cd browser-bridge
   ./install.sh

2. Load the extension once

   Open chrome://extensions, turn on Developer mode, choose Load unpacked and select the extension folder, then pin it. The badge turns into a purple dot when the agent connects.

3. Drive it

   chromectl open https://example.com
   chromectl snapshot
   chromectl click 12
   chromectl type 3 "hello" --submit
   chromectl screenshot