← Back to Discover

Cua

Open-source tooling that gives AI agents full desktops to work in: a desktop-driving layer, local VMs and sandboxes, SDKs, small decision models and a benchmark for computer-use agents.

Autonomous agentsBrowser automationAgent frameworks
Visit website ↗

What it does

Cua is a repository of components for computer-use agents, meaning agents that operate desktop apps and browsers. You bring your own agent and model, and Cua supplies the computer and the automation tools. Cua Spaces is a desktop app that gives agents full desktops, each one a Space: a macOS VM built locally on your Mac, or a Linux image, running on your Mac, on other machines you own or in your own AWS, Google Cloud or Modal account. Cua Driver lets an agent inspect and operate native apps and browsers on macOS, Windows and Linux through a CLI, MCP or typed SDKs. Lume creates local macOS and Linux VMs on Apple Silicon. The repository also holds an SDK and a cua command for sandboxes, CUA-S1 small models for computer-use decisions (an early research release with weights on Hugging Face), and Cua Bench for building tasks and evaluating agents. Most of the repository is MIT-licensed, while Cua Spaces is source-available under FSL-1.1-MIT.

Capabilities

  • Cua Spaces: agent desktops running as VMs on your Mac, your own machines or your cloud account
  • Cua Driver: operate native apps and browsers on macOS, Windows and Linux through CLI, MCP or SDKs
  • Lume: local macOS and Linux VMs on Apple Silicon
  • cua SDK and CLI for sandboxes, with Python, TypeScript and Swift packages
  • Cua Bench for creating computer-use tasks and evaluating agents

When to use it

  • You are building an agent that must use full desktop apps rather than only APIs
  • You need isolated macOS or Linux VMs for agents to run in
  • You want an environment and benchmark for testing computer-use agents