cua
SHA-256Open-source infrastructure for building, training, and deploying AI agents that control full desktops (macOS, Linux, Windows). Includes sandboxes, SDKs, and benchmarks.
Smart Download
Download Download Version
vcua-driver-v0.1.9 · 7 MB
Open-source toolkit for AI agents that control desktops, with sandboxes, SDKs, and benchmarks.
Core Features
- Background driver for macOS: control native apps without stealing cursor or focus
- Unified sandbox API: one interface for Linux/macOS/Windows/Android containers and VMs
- Built-in benchmarks: evaluate agents on OSWorld, ScreenSpot, and custom tasks
- CuaBot: co-op computer-use sandbox for coding agents with native GUI windows
- Lume: near-native macOS/Linux VM management on Apple Silicon
What It Can't Do
- •Cua is not a general-purpose desktop automation tool; it is designed for AI agent scenarios and requires an AI model to drive it. The Cua Driver background mode currently supports macOS only; Windows/Linux sandboxes require VM/container runtime. Local sandboxes using QEMU need virtualization support (macOS Hypervisor.framework, Linux KVM). CuaBot is early-stage and supports only selected coding agents.
Use Cases
- Train and evaluate computer-use AI agents for desktop automation
- Run GUI workflows in isolated sandboxes (browsers, design tools, game engines)
- Enable coding assistants (e.g., Claude Code) to perform click-and-type tasks in virtual desktops
Cua is an open-source platform that provides all the tools needed to create computer-use agents—AI programs that can see screens, click buttons, type text, and complete tasks autonomously across macOS, Linux, and Windows. It offers a unified API for sandboxes (local or cloud), a background driver that never steals your cursor, a CLI and MCP server for integrating with coding assistants like Claude Code, and a benchmark suite for evaluating agent performance. Unlike traditional UI automation frameworks (Playwright, Selenium) that focus on web automation, Cua is designed specifically for AI agents, supporting native desktop apps, canvas-based tools (Blender, Figma, game engines), and mobile gestures. Its key differentiators: (1) background computer-use on macOS without affecting user interaction, (2) one API for any OS (Linux, macOS, Windows, Android) and any runtime (container, VM, cloud).
Tags
Getting Started
Install the software
Double-click the downloaded installer and follow the prompts
Install Cua Driver: /bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/trycua/cua/main/libs/cua-driver/scripts/install.sh)"
Install Cua Python package: pip install cua
Create a sandbox and perform actions in Python, see docs at https://cua.ai/docs
- Install Cua Driver: /bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/trycua/cua/main/libs/cua-driver/scripts/install.sh)"
- Install Cua Python package: pip install cua
- Create a sandbox and perform actions in Python, see docs at https://cua.ai/docs
SHA-256 checksum verified
Checksum extracted from GitHub official Release page
SHA256 Checksum
f65081a40ae8e08b324b5dbe0132041680273097d15a10e57a63bbb228a13e59This checksum is extracted from the GitHub Release page. Verify file integrity after download.
All SHA-256 checksums on this platform are extracted from the project's official GitHub Release page, without any modification. You can independently verify them on the GitHub Releases page.
Open Source Transparency
View GitHub SourceUninstall Info
Uninstall Cua package: `pip uninstall cua`. Remove Cua Driver: run the installer's uninstall script or delete /usr/local/bin/cua-related files. For Lume VMs, use `lume vm stop` and `lume vm delete`.
No Extra Dependencies
Ready to use after download. No additional runtime required.
Having issues? Check the FAQ below
5 FAQs
Similar Projects
ollama
Ollama lets you download, run, and manage large language models locally. One command, multiple platforms, endless possibilities.
llama.cpp
High-performance LLM inference engine in C/C++ with minimal dependencies, supporting quantized models (1.5–8 bit) and diverse hardware (Apple Silicon, CUDA, Vulkan, etc.).
opencv
OpenCV is an open-source computer vision and machine learning library with over 2500 optimized algorithms for real-time image and video analysis.