Browser AI Agent

An open-source autonomous browser agent that can navigate websites, interact with web applications, and execute multi-step tasks using AI. This is the core browser automation engine that powers TaskCato.com - the AI co-founder platform for founders and startups.

🎯 What Is This?

This is a headless browser agent controlled by AI (LLM) that can:

Navigate websites autonomously
Fill out forms and interact with web UIs
Extract data and take screenshots
Execute multi-step workflows
Remember context across sessions
Run multiple browser instances simultaneously

Think of it as "Selenium meets ChatGPT" - but the AI decides what to do next.

🆚 Open Source vs TaskCato Premium

Feature	Open Source (This Repo)	TaskCato.com (SaaS)
Browser Automation	✅ Full access	✅ Enhanced
AI-Powered Navigation	✅ Yes	✅ Yes
Multi-Instance Support	✅ Up to 9 browsers	✅ Unlimited
Integrations	❌ None	✅ 15+ (HubSpot, Gmail, etc.)
Voice Calls	❌ No	✅ AI phone calls
Persistent Memory	✅ Basic	✅ Advanced company context
Team Collaboration	❌ No	✅ Yes
Support	Community	Priority
Hosting	Self-hosted	Cloud-hosted
Price	Free	$29/month

Use this open-source version if:

You want to build your own automation tools
You need browser automation for personal projects
You're a developer who wants to customize the AI behavior
You prefer self-hosting

Use TaskCato.com if:

You're a founder who needs end-to-end business automation
You want integrations with CRM, email, calendar, etc.
You need AI that can make phone calls and manage your entire workflow
You want a managed service with support

🚀 Quick Start

One-Click Installation

We've made it dead simple. No Python, no terminal commands, no configuration.

Mac Users

Download Browser Agent.app from Releases
Double-click to launch
That's it! The app handles everything automatically:
- Creates Python virtual environment
- Installs all dependencies
- Starts the server
- Opens the dashboard

Windows Users

Download Browser Agent.exe from Releases
Double-click to launch
Done! Same automatic setup as Mac

First Run

The app will show a loading screen while it sets up (first time only)
Dashboard opens automatically at http://localhost:8000
Click Settings to configure your LLM:
- LM Studio (recommended for privacy): Download from https://lmstudio.ai
- Gemini (cloud): Add your API key from https://aistudio.google.com
Start giving commands!

Example Commands

"Research the top 5 AI startups and take screenshots"
"Go to Hacker News and summarize the top 3 stories"
"Fill out the contact form on example.com"

🛠️ For Developers

Want to modify the code or run from source?

Prerequisites

Python 3.13+
Node.js 18+ (for Electron)
LM Studio or Gemini API key

Running from Source

# Clone the repository
git clone https://github.com/CloudCorpRecords/browserboi.git
cd browserboi

# Option 1: Run the Electron app (recommended)
cd launcher
npm install
npm start

# Option 2: Run server only (terminal-based)
python3 -m venv venv
source venv/bin/activate  # Windows: venv\Scripts\activate
pip install -r requirements.txt
playwright install chromium
python -m uvicorn browser_agent.server.main:app --host 0.0.0.0 --port 8000

Building the App

cd launcher
npm install
npm run build

# Outputs:
# Mac: dist/mac-arm64/Browser Agent.app
# Windows: dist/Browser Agent 1.0.0.exe

🏗️ Architecture

┌─────────────────────────────────────────┐
│  Electron App (Optional UI)             │
│  - Dashboard                             │
│  - 9-Grid Browser View                   │
│  - Real-time Logs                        │
└──────────────┬──────────────────────────┘
               │
┌──────────────▼──────────────────────────┐
│  FastAPI Server (Port 8000)              │
│  - WebSocket for real-time updates      │
│  - Multi-instance management (1-9)      │
│  - RESTful API                           │
└──────────────┬──────────────────────────┘
               │
┌──────────────▼──────────────────────────┐
│  AI Agent Core                           │
│  - LLM Integration (Gemini/LM Studio)   │
│  - Task Planning & Execution             │
│  - Memory Management                     │
└──────────────┬──────────────────────────┘
               │
┌──────────────▼──────────────────────────┐
│  Browser Manager (Playwright)            │
│  - Headless Chromium                     │
│  - Screenshot Capture                    │
│  - Page Interaction                      │
└─────────────────────────────────────────┘

📚 Features

Core Capabilities

Autonomous Navigation: AI decides where to click, what to type
Multi-Step Workflows: Chain complex actions together
Context Awareness: Remembers previous actions and decisions
Screenshot Analysis: AI "sees" the page and makes decisions
Text Extraction: Reads page content for better understanding
Session Persistence: Maintains cookies and login state

Advanced Features

9-Instance Grid: Run up to 9 browser agents simultaneously
Headless Mode: Runs in background without visible browser
CDP Support: Chrome DevTools Protocol for advanced control
Memory System: Stores user preferences and company context
Streaming Logs: Real-time visibility into agent actions

🛠️ Development

Project Structure

browser_agent/
├── core/
│   ├── agent.py          # Main AI agent logic
│   ├── browser.py        # Playwright browser manager
│   ├── llm.py           # LLM integration
│   └── memory.py        # Context & memory management
├── server/
│   ├── main.py          # FastAPI server
│   └── static/          # Web UI
└── tools/               # Agent tools (click, type, etc.)

launcher/                # Electron desktop app
├── main.js             # Electron main process
├── index.html          # Loading screen
└── package.json        # Dependencies

API Endpoints

Start a task:

curl -X POST http://localhost:8000/api/start/1 \
  -H "Content-Type: application/json" \
  -d '{"task":"Search for AI news on Hacker News"}'

Stop an agent:

curl -X POST http://localhost:8000/api/stop/1

Health check:

curl http://localhost:8000/api/diagnostic

WebSocket Events

Connect to ws://localhost:8000/ws/1 for real-time updates:

status - Agent status changes
log - Action logs
screenshot - New screenshots
error - Error messages

🤝 Contributing

We welcome contributions! This is an open-source project maintained by the TaskCato team.

How to Contribute

Fork the repository
Create a feature branch (git checkout -b feature/amazing-feature)
Commit your changes (git commit -m 'Add amazing feature')
Push to the branch (git push origin feature/amazing-feature)
Open a Pull Request

Areas We Need Help

🐛 Bug fixes and stability improvements
📝 Documentation and examples
🔌 New tool integrations
🧪 Test coverage
🎨 UI/UX improvements

📖 Use Cases

Web Scraping

"Go to ProductHunt and extract the top 10 products today"

Form Automation

"Fill out the contact form on example.com with my info"

Research

"Research the pricing pages of Stripe, Square, and PayPal"

Testing

"Test the login flow on staging.myapp.com"

Data Collection

"Find email addresses of CTOs at Series A fintech startups"

⚙️ Configuration

Environment Variables

# LLM Provider
LLM_PROVIDER=lm_studio  # or 'gemini'
GEMINI_API_KEY=your_key_here

# Browser Settings
HEADLESS=true  # Run browser in headless mode
VIEWPORT_WIDTH=1280
VIEWPORT_HEIGHT=800

# Server
PORT=8000
HOST=0.0.0.0

Settings File

Settings are stored in browser_agent/data/settings.json:

{
  "provider": "lm_studio",
  "lm_studio_url": "http://localhost:1234/v1",
  "lm_studio_model": "qwen/qwen3-vl-8b",
  "gemini_api_key": "",
  "gemini_model": "gemini-2.0-flash-exp"
}

🔒 Security & Privacy

No Data Collection: All data stays on your machine
Local LLM Support: Use LM Studio for complete privacy
Session Isolation: Each browser instance is isolated
No Telemetry: We don't track usage

🐛 Known Issues

UI Message Display: Agent responses may not appear in chat (check terminal logs)
Browser View Content: Grid cells show blank instead of live browser content
Windows Support: Electron app tested primarily on macOS

See Issues for full list.

📄 License

MIT License - see LICENSE file for details.

🌟 About TaskCato

This browser agent is the core technology behind TaskCato.com - an AI co-founder platform that helps founders automate their entire business workflow.

TaskCato adds:

15+ SaaS integrations (HubSpot, Gmail, Slack, etc.)
AI voice calling capabilities
Advanced persistent memory
Team collaboration features
Managed cloud hosting
Priority support

Try TaskCato free: https://taskcato.com

💬 Community

Discord: Join our community
Twitter: @TaskCato
Issues: GitHub Issues
Discussions: GitHub Discussions

🙏 Acknowledgments

Built with:

Playwright - Browser automation
FastAPI - Web framework
Electron - Desktop app
Google Gemini - AI capabilities
LM Studio - Local LLM support

Made with ❤️ by the TaskCato team

Building the future of autonomous work, one browser action at a time.

Name		Name	Last commit message	Last commit date
browser_agent		browser_agent
launcher		launcher
.gitignore		.gitignore
Launch.command		Launch.command
README.md		README.md

Latest commit

History

Folders and files

Repository files navigation