Gemini in Chrome

A CLI coding agent powered by the Gemini API with full Chrome browser automation via the Chrome DevTools Protocol.

Install

git clone https://github.com/Beandon13/geminiInChrome.git
cd geminiInChrome
npm install
npm run build
npm link

This installs the global gemini command.

Setup

Run gemini for the first time and the setup wizard will walk you through:

1. API Key — Get one free at Google AI Studio
2. Model — Pick from gemini-2.5-flash, gemini-2.5-pro, or gemini-2.0-flash
3. Chrome port — Port for Chrome remote debugging (default: 9222)

Config is saved to ~/.gemini-chrome/config.env. Run gemini --setup to change settings later.

Chrome Setup

To use browser automation features, launch Chrome with remote debugging:

/Applications/Google\ Chrome.app/Contents/MacOS/Google\ Chrome --remote-debugging-port=9222

On Linux:

google-chrome --remote-debugging-port=9222

Usage

gemini                              # Start a new session
gemini --dir ~/projects/my-app      # Work in a specific directory
gemini --resume 2025-01-15_14-30    # Resume a previous session
gemini --setup                      # Re-run setup wizard

CLI Commands

Command	Description
/tabs	List Chrome tabs
/attach N	Attach to tab N
/cd <path>	Change working directory
/clear	Clear screen
/status	Session info
/exit	Quit
Ctrl+C	Interrupt current task
Ctrl+C x2	Exit

What It Can Do

Coding — Read, write, edit, and search files. Run shell commands.

Browser — Navigate pages, click elements, type text, take screenshots, scroll, read page content.

Just describe what you want in plain English:

> Fix the bug in src/utils.ts where date parsing fails
> Go to github.com and star the repo
> Read the error on the page and suggest a fix

Requirements

• Node.js 18+
• Google Chrome (for browser features)
• Gemini API key