🧠 Agent OS — Autonomous Multi-Model AI Workspace

📸 Screenshots

💬 Chat + Live Preview Workspace

Split-pane layout: chat on the left, live app preview on the right. Attach files, take photos, or use the camera — all from the input bar.

⚙️ Multi-Provider Settings

Manage API keys for 7 AI providers. Green dot = connected, red dot = needs key. Add any custom model ID.

🖥️ Local Models & Custom Configuration

Ollama and LM Studio for local models (no API key needed). Add custom models from any provider. Override the system prompt.

✨ Features

🤖 Autonomous Agent

• ReAct Loop — Thinks, acts, observes, and iterates autonomously (up to 15 steps)
• 9 Built-in Tools — Search, browse web, write/read files, run Python, execute shell commands
• Any Task — "Create a todo app", "Search the web", "Run Python code", "Build a calculator"

🎨 Live Preview (Claude Artifacts-Style)

• Split Layout — Chat on the left, live preview on the right
• Instant Rendering — Created apps appear immediately in an interactive iframe
• Code Viewer — View source code of any created file
• Output Panel — See tool execution logs in real-time

📎 File & Camera Support

• Attach Files — Upload any file type to the workspace
• Photo Upload — Attach images from your device
• Camera Capture — Take photos directly from your webcam
• Drag & Drop — Drop files into the chat area

⚙️ Multi-Provider API Management

• 7 AI Providers with dedicated settings cards
• Per-provider API keys — saved to .env for persistence
• Custom Models — Add any model ID from any provider
• Local Models — Ollama & LM Studio support (no API key needed)

🚀 Quick Start

Prerequisites

• Python 3.10+
• An API key from any supported provider

Installation

# Clone the repository
git clone https://github.com/Aj2280/Agent-OS.git
cd Agent-OS

# Install dependencies
pip install -r requirements.txt

# Set your API key (or add it later in Settings)
echo "OPENROUTER_API_KEY=your-key-here" > .env

# Start the server
python server.py

Open in Browser

http://localhost:8000

That's it! 🎉

🔌 Supported Providers

Tip: OpenRouter is the easiest way to get started — one key gives you access to 100+ models.

🛠️ Built-in Tools

🏗️ Architecture

Agent-OS/
├── server.py          # FastAPI backend — API routes, file serving
├── agent_core.py      # Autonomous ReAct agent with multi-provider routing
├── tools.py           # 9 tool implementations
├── main.py            # CLI interface (optional)
├── requirements.txt   # Python dependencies
├── .env               # API keys (auto-generated)
├── static/
│   ├── index.html     # Dashboard UI
│   ├── style.css      # Dark theme, glassmorphism
│   └── app.js         # Frontend logic, preview, uploads
└── workspace/         # Files created by the agent

How It Works

graph LR
    A[User Message] --> B[FastAPI Server]
    B --> C[Agent Core]
    C --> D{LiteLLM Router}
    D --> E[OpenRouter/OpenAI/Gemini/...]
    C --> F[Tool Execution]
    F --> G[Write File / Run Code / Search]
    G --> C
    C --> H[Response + Created Files]
    H --> I[Live Preview in iframe]

1. User sends a message via the chat UI
2. Agent plans using the ReAct loop (Reason → Act → Observe)
3. Tools are executed — files written, code run, web searched
4. Results are returned — response in chat, apps in live preview
5. Files are served from the workspace via /preview/ endpoint

📦 Dependencies

fastapi
uvicorn
litellm
duckduckgo-search
requests
beautifulsoup4
python-dotenv
python-multipart

🤝 Contributing

Contributions are welcome! Feel free to:

• Open issues for bugs or feature requests
• Submit pull requests with improvements
• Add new tools or providers

📄 License

MIT License — feel free to use, modify, and distribute.