About Rebrowse

Rebrowse is a powerful tool that converts screen recordings into automated workflow agents.

Key features of Rebrowse:

• Record your screen once and automate the workflow forever
• Transform recordings into shareable workflow agents
• Support for cross-application workflows
• Easy sharing and collaboration

Explore our Showcases at rebrowse.me

Watch Demo

Click and watch Youtube videos.

▶ Lead Gen demo

▶ Remove background from Rick Rubin image

▶ Tweet post

Follow me on Twitter for the latest updates!

---

Two Agent Modes

✍🏻 Text prompt

• 👍 Robust.
• 👎 needs long descriptive prompt. slow.

▶️ Recording

• 👍 Determinictic. 10x fast. 3x Accurate.
• ⚠️ AI-powered fall-back is WIP.

---

Installation Guide

Prerequisites

• Python 3.8 or higher
• Git
• Chrome browser (for browser automation)
• Gradio (for the web interface)

Step 1: Clone the Repository

git clone https://github.com/zk1tty/rebrowse-app.git
cd rebrowse-app

Step 2: Set Up Python Environment

1. case: MacOS

   # Install uv:
   curl -LsSf https://astral.sh/uv/install.sh | sh
   
   # Create and activate a virtual environment
   uv venv --python 3.11
   
   # activate venv:
   source .venv/bin/activate
   
   # Install python dependencies
   uv pip install -r requirements.txt

2. case: Windows

   # Install uv:
     powershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/install.ps1 | iex"
   
   # Create and activate a virtual environment
   uv venv --python 3.11
   
   # activate venv:
   .venv\Scripts\activate
   
   # Install python dependencies
   uv pip install -r requirements.txt

Step 3: Configure Environment Variables

1. Create a new Chrome profile to be used by rebrowse

   mkdir -p custom_chrome_profile
   

2. Copy the .env.example file to a new file named .env in the project root.

   cp .env.example .env
   

3. Open .env file

   vim .env
   

4. add API key and chrome paths

   # LLM API Keys
   OPENAI_API_KEY=your_openai_api_key

   # - Windows: `C:\Program Files\Google\Chrome\Application\chrome.exe`
   # - Linux: `/usr/bin/google-chrome`
   CHROME_PATH="/Applications/Google Chrome.app/Contents/MacOS/Google Chrome"
   
   # ⚠️ Replace {username}
   CHROME_USER_DATA="/Users/{username}/Projects/rebrowse-app/custom_chrome_profile"
   CHROME_CDP_URL="http://localhost:9222"

Step 4: Close all open Chrome tabs. IMPORTANT!!

Or, run this command.

pkill -9 "Google Chrome"

Step 5: Launch the Web Application

# Start with Gradio
gradio webui.py

# Or run directly with Python
python webui.py

The application will be available at http://127.0.0.1:7860

Step 6: Use FireFox or Safari to open 127.0.0.1:7860

• DO NOT use Chrome to open this web app.
• Open Safari or FIrefox, and go to "http://127.0.0.1:7860"

Step 7: Run a clean Chrome with CDP session

⚠️ I am temporarily making the test profile on the same project file, and name it /custom_chrome_profile for debugging purposes.

1. Run this Chrome using the new profile. Configuration details to connect to your own Chrome are here.

   sh run_custom_chrome.sh

2. Make sure that the DevTools WebSocket session is available

   DevTools listening on ws://127.0.0.1:9222/devtools/browser/1bbd94a9-aed9-4462-bc25-fddec9d9663c
   

3. Log in to web apps on the new Chrome profile In this process, you will create a new Chrome profile to be used by browser-agents.

   • log in to your web accounts: X, LinkedIn, Youtube, etc.
   • If the browser agent can skip this process, it will be easier to handle executions.

Step 8: Ready to play ^^

• Go to "Choose Agent" tab
• Pick up a Preset Task
• Try "Run Agent"
• Enjoy your browsing agent.

---

Are you a dev?

Let me navigate you through the technical concept and objectives of Rebrowse.

Background

This project builds upon the foundation of browser-use, which is designed to make websites accessible for AI agents.

The original creator is WarmShao, who made the WebUI.

Key difference from Browser-use?

I needed to make the workflow of agent behaibier by faster and accurate.
One approach to do this is recoeding and make it repeatable.
I introduced our technical architecture in Architecure.

Let me share our technical milestones.

• Make a workflow traceable
• Implement Replay mode with Replayer
• Add an eval process to trigger Drift or not. <- enough flexible
• Take a demo video and add here.
• Measure the accuracy and speed of replay mode in comparison with agent mode
• Add Analyzer: a reasoning process to make the process modifiable by test prompt
• Add Remixer
• Design Replayer memory to handle edge cases
• Add multi-thread execution of a single workflow on Replayer

Update on 19th May

I found that @browser-use team released workflow-use. I'm still researching their approach and objectives.
Let's talk more on X or Telegram.

Product Roadmap

We are going to build a marketplace, where users can share cross-app workflows by recording, instead of node editors like Zapier or n8n.
Check out our Roadmap.

Gradio Development Mode

For development with auto-reload:

gradio webui.py --watch src

This will automatically reload the browser when you make changes to files in the src directory.