ultron-ai
ULTRON AI - Open Source Manus AI Alternative. General-purpose AI Agent with LM Studio local model support.
? ULTRON AI
Open-Source Autonomous AI Agent Platform
Built by Maharab Hossen
Quick Start • Features • Architecture • Configuration • ????
?? What is ULTRON AI?
ULTRON AI is an open-source, general-purpose AI agent system that can autonomously plan, reason and execute complex tasks. Give it a goal and it will break it down, use tools, browse the web, write code, manipulate files, and deliver results — all inside a secure sandboxed environment.
Think of it as your own self-hosted alternative to Manus AI, ChatGPT with Code Interpreter, or Devin — but fully open source and under your control.
How ULTRON AI Compares
| Capability | ULTRON AI | Manus AI | ChatGPT |
|---|---|---|---|
| Open Source | ? | ? | ? |
| Self-Hosted | ? | ? | ? |
| Web Browsing | ? | ? | ? |
| Code Execution | ? | ? | ? |
| File Operations | ? | ? | ? |
| Sandboxed Environment | ? | ? | ? |
| MCP Tool Integration | ? | ? | ? |
| Custom LLM Backend | ? | ? | ? |
| Local / Offline Mode | ? | ? | ? |
| Free to Use | ? | ? | ? |
? Features
- ?? Autonomous Planning — Breaks complex tasks into actionable steps using Plan-Act architecture
- ?? Live Web Browser — Full Chrome browser running inside the sandbox, viewable via NoVNC in real-time
- ?? Shell & Code Execution — Runs bash commands, Python scripts, and more inside an isolated Ubuntu container
- ?? File Management — Creates, reads, edits, and organizes files within the sandbox filesystem
- ?? MCP Tool Ecosystem — Extend capabilities with Model Context Protocol servers (GitHub, databases, APIs, etc.)
- ?? Web Search — Integrated search via Bing, Google, or Baidu
- ?? Secure Sandbox — Every task runs in a disposable Docker container with automatic cleanup
- ?? Multi-User Auth — Built-in authentication with email/password, local admin, or no-auth modes
- ?? Modern Web UI — Clean Vue 3 interface with real-time streaming, session sharing, and tool visualization
- ?? Any LLM — Works with OpenAI, DeepSeek, LM Studio, Ollama, or any OpenAI-compatible API
?? Quick Start
Prerequisites
- Docker Desktop (Windows/Mac) or Docker Engine (Linux) — v20.10+
- An LLM with function calling support (OpenAI, DeepSeek, LM Studio, etc.)
1. Clone & Configure
git clone https://github.com/maharab549/ultron-ai.git
cd ai-manus
cp docker-compose-example.yml docker-compose.yml
Edit docker-compose.yml and set your LLM provider:
OpenAI:
- API_BASE=https://api.openai.com/v1
- API_KEY=sk-your-key-here
- MODEL_NAME=gpt-4o
DeepSeek:
- API_BASE=https://api.deepseek.com/v1
- API_KEY=sk-your-key-here
- MODEL_NAME=deepseek-chat
LM Studio (Local):
- API_BASE=http://host.docker.internal:1234/v1
- API_KEY=lm-studio
- MODEL_NAME=your-model-name
Ollama (Local):
- API_BASE=http://host.docker.internal:11434/v1
- API_KEY=ollama
- MODEL_NAME=llama3
2. Launch
docker compose up -d
3. Open
Visit http://localhost:5173 in your browser. That’s it!
Note: You’ll see
sandbox-1 exited with code 0— this is expected. It pre-pulls the sandbox image.
?? Architecture
ULTRON AI follows a clean, modular architecture:
+-------------+ +--------------+ +-----------------+
¦ Frontend ¦----?¦ Backend ¦----?¦ Sandbox ¦
¦ (Vue 3) ¦?----¦ (FastAPI) ¦?----¦ (Ubuntu+Chrome)¦
+-------------+ SSE +--------------+ +-----------------+
¦ ¦
+---------+ +-----------+
¦ MongoDB ¦ ¦ Browser ¦
¦ Redis ¦ ¦ Shell ¦
+---------+ ¦ Files ¦
+-----------+
How it works:
- User sends a task through the web interface
- Backend creates a fresh Docker sandbox and routes the task to the PlanAct Agent
- The agent decomposes the task into steps, calling tools (browser, shell, files, search, MCP) as needed
- All progress streams back to the frontend via Server-Sent Events (SSE)
- User can watch the browser in real-time via NoVNC and inspect all tool outputs
?? Configuration
| Variable | Description | Default |
|---|---|---|
API_BASE |
LLM API endpoint | http://mockserver:8090/v1 |
API_KEY |
LLM API key | — |
MODEL_NAME |
Model identifier | deepseek-chat |
TEMPERATURE |
Response randomness (0-1) | 0.7 |
MAX_TOKENS |
Max response tokens | 2000 |
SEARCH_PROVIDER |
Search engine (bing, google, baidu) |
baidu |
AUTH_PROVIDER |
Auth mode (password, local, none) |
password |
MCP_CONFIG_PATH |
Path to MCP config file | /etc/mcp.json |
LOG_LEVEL |
Logging level | INFO |
See full configuration docs for all options including MongoDB, Redis, JWT, email, and sandbox settings.
??? Development
For local development with hot-reload:
# Clone
git clone https://github.com/maharab549/ultron-ai.git
cd ai-manus
# Copy env template
cp .env.example .env
# Edit .env with your LLM settings
# Start all services
./dev.sh up -d
# View logs
./dev.sh logs -f backend
See LOCAL_SETUP_GUIDE.md for detailed development setup instructions.
?? Project Structure
ultron-ai/
+-- frontend/ # Vue 3 + TypeScript web interface
+-- backend/ # FastAPI Python backend (DDD architecture)
¦ +-- app/
¦ +-- domain/ # Business logic, agents, flows
¦ +-- infrastructure/# LLM clients, DB, external services
¦ +-- interfaces/ # API routes, WebSocket handlers
¦ +-- application/ # Application services
+-- sandbox/ # Ubuntu Docker sandbox with Chrome + tools
+-- mockserver/ # Development mock LLM server
+-- docs/ # Documentation site (Docsify)
+-- docker-compose.yml # Production deployment
??? Roadmap
?? Contributing
Contributions are welcome! Feel free to:
- 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
?? License
This project is open-source. See the LICENSE file for details.
? Star this repo if you find ULTRON AI useful!
https://github.com/maharab549/ultron-ai