cuse

An open-source framework for building AI agents that can interact with computers

Features

• Computer Control: Display, mouse, and keyboard interaction
• Authentication: Authenticate with credentials
• File Operations: View, create, and edit files
• Shell Access: Execute commands and manage processes
• App Framework: Build custom applications
• Linux Support: Run via Docker containers

Demo

Task: Check gmail inbox and send summary to {email}.

cuse.demo.mp4

Quickstart

Install dependencies:

npm install @cusedev/core

Initialize and create a computer:

npx @cusedev/cli init

Create a Computer instance

import { Computer } from '@cusedev/core';

const computer = new Computer();

Interact with the computer:

// Take a screenshot
const screenshot = await computer.system.display.getScreenshot();

// Type some text
await computer.system.keyboard.type({ text: 'Hello, World!' });

// Execute a command
const output = await computer.system.bash.execute({ command: 'ls -la' });

Documentation

Visit our documentation to learn more about:

• Getting started with the example project
• Adding cuse to your existing project
• Core concepts and API reference
• CLI commands and usage

Roadmap

• Support for other platforms
• Deployment
• Stateful Machines
• Reusable Workflows

Contributing

Contributions are welcome! Please check out our GitHub repository.

License

MIT License — see LICENSE file

Get in Touch

• Visit our website
• Join our Discord community

Star History