Aranya Dutta
Writing/BrowserBridge
2026.03.11Projects

BrowserBridge

Personal Notes

Hey, so basically I was working on making an AI automation where I needed the AI agent to be able to use my real browser session. But I couldn’t use typical automation tools like Puppeteer, Selenium, or headless browsers, because they run in separate browser instances and don’t easily reuse my existing login session, cookies, and browser state.

So I thought, why not make a minimal setup that allows me to control and use basic features of browsing programmatically from outside the browser. That’s how this project started. The initial goal was just to create a small bridge between external tools (like AI agents or automation workflows) and my actual browser session.

Over time, the idea evolved a bit and eventually turned into the current setup you see here.

If you have a better solution for this problem, or ideas to improve this project, definitely let me know! I’d love to hear it.

And yeah, of course you guessed it right… it’s 2026+ (and maybe even later when you’re reading this), so I explained the project to an AI and let it help write the README 😂. But hey, it’s still pretty good. Have fun! 😼🤓

Project Link: https://github.com/thisizaro/BrowserBridge

Introduction

BrowserBridge is a local browser automation interface that allows external systems such as AI agents, scripts, or workflow tools to control a real web browser through WebSocket commands.

Unlike traditional automation tools, BrowserBridge operates inside the user’s actual browser session. This means it preserves:

  • login sessions
  • cookies
  • local storage
  • dynamically rendered pages

This makes it suitable for automating authenticated websites and modern JavaScript applications.

Architecture

Controller (AI / scripts / n8n) communicates with the browser through a FastAPI WebSocket router.

Controller
    │
    │ WebSocket
    ▼
FastAPI Server
    │
    │ WebSocket
    ▼
Browser Extension
    │
    ▼
Browser Tabs / DOM

Project Structure

browserbridge/
│
├── server/
│   └── main.py
│
├── extension/
│   ├── manifest.json
│   ├── background.js
│   └── content.js
│
├── client_test.py
├── requirements.txt
│
└── docs/
    └── commands.md

Setup

1. Start the Python Server

Create and activate a virtual environment:

python -m venv venv

Activate it.

Linux / Mac

source venv/bin/activate

Windows

venv\Scripts\activate

Install dependencies:

pip install -r requirements.txt

Run the FastAPI server:

uvicorn main:app --reload --port 8000

WebSocket endpoints:

ws://localhost:8000/ws/browser
ws://localhost:8000/ws/controller

2. Install the Browser Extension

  1. Open the Chrome extensions page:
chrome://extensions

  1. Enable Developer Mode

  2. Click Load unpacked

  3. Select the extension folder

The extension will automatically connect to the server.

3. Test the System

Run the test controller:

python client_test.py

Example command sent to the browser:

{
  "type": "command",
  "id": "cmd_001",
  "action": "open_url",
  "params": {
    "url": "https://example.com"
  }
}

This will open the URL in a new browser tab.

Documentation

Detailed command documentation is available in:

docs/commands.md

Current Commands

  • open_url
  • get_html
  • eval_js

Additional commands will be added in future versions.

Roadmap

Upcoming improvements:

  • command queue
  • tab targeting
  • page load waiting
  • DOM query commands
  • browser interaction commands
  • AI automation support

Status

This project is currently in early development.

automationbrowserai-agents