cua and Computer-Use-Agent

The open-source infrastructure providing sandboxes and SDKs for training computer-use agents (A) serves as the foundational platform that a specific agent implementation (B) would build upon or be evaluated against.

cua
69
Established
Computer-Use-Agent
21
Experimental
Maintenance 25/25
Adoption 10/25
Maturity 16/25
Community 18/25
Maintenance 6/25
Adoption 6/25
Maturity 9/25
Community 0/25
Stars: 13,043
Forks: 805
Downloads:
Commits (30d): 108
Language: Python
License: MIT
Stars: 16
Forks:
Downloads:
Commits (30d): 0
Language: Jupyter Notebook
License: Apache-2.0
No Package No Dependents
No Package No Dependents

About cua

trycua/cua

Open-source infrastructure for Computer-Use Agents. Sandboxes, SDKs, and benchmarks to train and evaluate AI agents that can control full desktops (macOS, Linux, Windows).

Provides unified APIs for sandbox creation, UI automation (screenshots, mouse/keyboard/touch), and code execution across heterogeneous environments (containers, VMs, cloud, local QEMU) with platform-specific optimizations like H.265 streaming and native window integration. Includes cua-bench for standardized evaluation on OSWorld and ScreenSpot, plus Lume for high-performance macOS virtualization on Apple Silicon via Virtualization.Framework.

About Computer-Use-Agent

Codeeaner/Computer-Use-Agent

An AI Agent that is able to control your screen to complste any task

Implements a local, closed-loop automation system using Qwen3-VL vision-language model via Ollama to iteratively capture screenshots, analyze UI state, and execute mouse/keyboard actions on Windows 11. Built with `mss` for efficient screen capture and `pyautogui` for input control, it includes function calling for structured action planning and screenshot history for debugging automation workflows.

Scores updated daily from GitHub, PyPI, and npm data. How scores work