cua
Scale computer-use 2.0 with open-source drivers, cross-OS fleets, and benchmarks for training, evaluation, and data generation.
Cua is an open-source computer-use agent skill that enables AI agents to control desktop environments across macOS, Windows, Linux, and Android. It provides background automation drivers that let agents click, type, and verify UI elements without interrupting the user's cursor or focus, plus cross-platform sandboxes for safe task execution. The platform includes an MCP server for integration with Claude Code, Cursor, and other AI coding assistants, alongside benchmarking tools for evaluation and training. Developers can automate native desktop applications, run agent tasks in isolated environments, and scale computer-use workflows across multiple operating systems.
Key Features
Use Cases
- 01Building AI coding agents that automate desktop application testing across multiple operating systems
- 02Enabling Claude Code or Cursor to interact with native GUI applications in the background during development
- 03Creating autonomous agents that complete multi-step desktop workflows without user intervention
- 04Evaluating computer-use agent performance using standardized benchmarks like OSWorld and ScreenSpot
- 05Training reinforcement learning models on desktop interaction tasks with exported trajectory data
- 06Running isolated macOS VMs on Apple Silicon for agent testing and development
Related Skills
View moresuperpowers
An agentic skills framework & software development methodology that works.
skills
Skills for Real Engineers. Straight from my .agents directory.
skills
Public repository for Agent Skills
ponytail
Makes your AI agent think like the laziest senior dev in the room. The best code is the code you never wrote.
cua — FAQ
What is Cua?+
Cua is an open-source platform for building AI agents that control computers through screen interaction, clicks, and typing. It includes background automation drivers, cross-OS sandboxes, an MCP server for AI coding assistants, and benchmarking tools for training and evaluation.
How do I install the Cua driver for macOS or Linux?+
Run the install script with `/bin/bash -c "$(curl -fsSL https://cua.ai/driver/install.sh)"` and follow the post-install instructions. For Windows, use PowerShell: `irm https://cua.ai/driver/install.ps1 | iex`.
Which AI clients work with Cua's MCP server?+
Cua's MCP server integrates with Claude Code, Cursor, Codex, OpenClaw, and other custom clients that support the Model Context Protocol.
Do I need API keys or credentials to use Cua?+
For local sandbox usage with QEMU, no API keys are required. Cloud sandboxes via cua.ai require an account. The driver and MCP server work locally without credentials.
Is Cua free to use?+
Yes, Cua is open-source under the MIT License. Local drivers, sandboxes, and benchmarks are free; cloud-hosted sandboxes on cua.ai may have separate pricing.
What are the prerequisites for running Cua sandboxes?+
Python 3.11 or later is required. Local sandboxes need QEMU installed; cloud sandboxes work through the cua.ai service without local virtualization.
How do I install cua?+
Open the source repository on GitHub and follow its README. cua is a skill — MCP Agents Market links you directly to the official repo.
Is cua free?+
cua is an open-source project hosted on GitHub. Check the repository for its license and any usage requirements.