OpenDesktop -- AI-Powered Desktop Agent

NOT A CHATBOT. AN INTELLIGENCE AGENT.

OpenDesktop

A self-improving, self-rewriting AI desktop agent that controls your computer, researches the web, spawns AI armies, trains custom models, and gets smarter every single interaction. 44 subsystems. 37 modules. 50+ AI providers.

terminal -- opendesktop

// Capabilities

Everything your computer can do. Now autonomous.

OpenDesktop is not a chatbot with a terminal plugin. It is a full intelligence agent with deep OS integration, visual understanding, and the ability to rewrite its own code.

Desktop Control

Mouse, keyboard, window management, file operations, clipboard intelligence, process monitoring. Full OS integration.

[ ]

Code Intelligence

37 programming languages. Auto detection, project scaffolding, code review, debugging, test generation. Write and execute code in any language.

{ }

Sub-Agent Army

Spawn hundreds of specialized sub-agents. 8 specializations. Team-based parallel execution. Shared knowledge between agents.

( )

Visual Understanding

Screen analysis and OCR. Find and click UI elements by description. Watch mode for continuous monitoring. Read everything on screen.

Self-Evolution

Self-code rewriting. Correction learning. Skill crystallization. A/B testing framework. Automatic improvement suggestions. Gets better every interaction.

Knowledge Graph Brain

5 memory types. Weighted nodes with auto-relating and decay. Full-text search. Persistent across sessions. Export and import.

Web Research

Real web search across DuckDuckGo, Wikipedia, StackOverflow, GitHub, npm. Deep multi-source research. URL scraping. Browser automation.

IoT Control

Philips Hue, LIFX, Tasmota, Shelly, Home Assistant. MQTT support. Device discovery. Automation rules. Control your smart home.

20 Messaging Platforms

Telegram, Discord, WhatsApp, Slack, Signal, iMessage, IRC, Matrix, LINE, Viber, MS Teams, Twitch, Email, SMS, and more.

// Installation

Up and running in seconds.

One command. That is it. Works on Linux, macOS, and Windows.

Windows (PowerShell) RECOMMENDED

# One-line install (handles Node.js too)
powershell -ExecutionPolicy Bypass \
-c "irm https://raw.githubusercontent.com/
Atum246/OpenDesktop/main/install.ps1 | iex"

Windows (cmd) CMD

# Download and run installer
curl -o install.bat \
https://raw.githubusercontent.com/
Atum246/OpenDesktop/main/install.bat
install.bat

Linux / macOS UNIX

# Install via curl
curl -fsSL \
https://raw.githubusercontent.com/
Atum246/OpenDesktop/main/install.sh | bash

npm (Any OS) NPM

# Install globally via npm
npm install -g opendesktop-ai

# Run setup wizard
opendesktop --setup

# Launch
opendesktop

Docker DOCKER

# Build the image
docker build -t opendesktop .

# Run the container
docker run -it \
  -p 4444:4444 \
  -e OPENDESKTOP_API_KEY=your-key \
  opendesktop

Requirements SYSTEM

# Minimum requirements
Node.js      v18.0.0 (v20+ recommended)
npm          v8.0.0 (v10+ recommended)
RAM          4GB    (8GB+ recommended)
OS           Linux, macOS 12+, Win 10+
Network      Internet connection

// AI Providers

50+ providers. One interface. Zero lock-in.

Use any model from any provider. Switch with a single command. No code changes required.

OpenRouter

200+ models

OpenAI

GPT-4o, o1

Anthropic

Claude 3.5

Google AI

Gemini Pro

Groq

Llama 3.1

DeepSeek

Chat, Coder

xAI

Grok 2

Mistral

Large, Mixtral

Cohere

Command R+

Perplexity

Sonar

01.AI

Yi Large

Moonshot

128k context

Zhipu AI

GLM-4

Baidu

ERNIE 4.0

Alibaba

Qwen Max

MiniMax

Abab 6.5

// Architecture

Modular by design. Extensible by nature.

37 independent modules orchestrated by a central engine. Every subsystem is isolated, testable, and replaceable.

project structure

OpenDesktop/

├── bin/opendesktop # CLI entry point

├── src/

│ ├── core/ # Engine + config (44 subsystems)

│ ├── providers/ # 50+ AI providers

│ ├── brain/ # Knowledge graph brain

│ ├── os-integration/ # Deep OS control

│ ├── evolution/ # Self-evolution system

│ ├── sub-agents/ # Agent orchestrator

│ ├── security/ # AES-256, audit logging

│ ├── visual-understanding/ # Screen vision + OCR

│ ├── iot/ # Smart home control

│ ├── messaging/ # 20 chat platforms

│ └── ... # 25 more modules

├── tests/test.js # 119 tests

├── Dockerfile # Docker support

└── package.json

// Commands

Full control from the terminal.

Over 100 slash commands for every subsystem. Type /help to see them all.

Desktop

/screenScreenshot and analyze

/clickClick element by description

/openOpen application

/browseOpen URL in browser

/processesList running processes

/lockLock screen

Code

/codeExecute code (37 languages)

/projectScaffold new project

/deployDeploy to 18 targets

/code-reviewReview code file

/debugDebug error message

/generate-testsGenerate test suite

Intelligence

/orchestrateSpawn agent team

/deep-searchMulti-source research

/evolveTrigger self-evolution

/brainKnowledge graph status

/proactiveProactive insights

/correctTeach correction

System

/backupBackup management

/scheduleTask scheduler

/monitorPerformance metrics

/marketplacePlugin marketplace

/trustTrust and safety

/profilesConfig profiles