Agent-OS

agent
Security Audit
Warn
Health Warn
  • License — License: MIT
  • Description — Repository has a description
  • Active repo — Last push 0 days ago
  • Low visibility — Only 6 GitHub stars
Code Warn
  • network request — Outbound network request in static/app.js
Permissions Pass
  • Permissions — No dangerous permissions requested

No AI report is available for this listing yet.

SUMMARY

Autonomous Multi-Model AI Workspace - Claude Artifacts-style live preview, 7 AI providers, file/camera uploads, 9 built-in tools

README.md

🧠 Agent OS — Autonomous Multi-Model AI Workspace

Agent OS Banner
Python
FastAPI
License

A Claude Workspace-like autonomous AI agent that creates real apps, runs code, and shows live previews — all from your browser.

FeaturesScreenshotsQuick StartProvidersArchitecture


📸 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.

Chat Dashboard

⚙️ Multi-Provider Settings

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

Settings Panel

🖥️ 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.

Settings Bottom


✨ 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

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

Provider Models API Key Required
🌐 OpenRouter 100+ models (GPT-4o, Claude, Llama, etc.)
🤖 OpenAI GPT-4o, GPT-4 Turbo, GPT-3.5
💎 Google Gemini Gemini 2.0 Flash, 1.5 Pro
🧬 Anthropic Claude 3.5 Sonnet, Claude 3 Opus
Groq Llama 3.1, Mixtral (ultra-fast)
🏠 Ollama Any local model (Llama, Mistral, etc.)
🖥️ LM Studio Any GGUF model loaded locally

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


🛠️ Built-in Tools

Tool Description
google_search Search the web for real-time information
fetch_url Read and extract content from web pages
write_file Create files (code, HTML, documents)
read_file Read existing files in the workspace
list_files Browse directory contents
create_directory Create folders
run_python_code Execute Python code directly
run_shell_command Run system commands
think Internal planning for complex tasks

🏗️ 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.


Built with ❤️ by Aj2280

⭐ Star this repo if you find it useful!

Reviews (0)

No results found