trycua/cua

Scale computer-use 2.0 with open-source drivers, cross-OS fleets, and benchmarks for training, evaluation, and data generation.

What it solves

Cua provides the infrastructure and tools necessary for AI agents to interact with computers. It solves the challenge of giving agents a safe, isolated environment (sandboxes) and the ability to operate native desktop applications across macOS, Windows, and Linux, rather than being limited to APIs or web browsers.

How it works

Cua operates through several integrated components:

  • Cua Driver: A tool that allows agents to inspect and operate native desktop apps and browsers. It supports background delivery, allowing agents to work without taking focus or moving the user's pointer.
  • Sandboxes & Fleets: Provides isolated cloud desktops (Fleets) or local containers/VMs (Lume) that agents can use as their workspace.
  • CUA-S1: A family of small, specialized "System 1" models designed for fast, bounded decisions, such as filling out forms or interacting with interface elements.
  • Cua Bench: A framework for building tasks, evaluating agent performance, and exporting trajectories for training.
  • SDKs & CLI: A unified interface (Python, TypeScript, Swift, Kotlin) to manage these environments and tools.

Who it’s for

Developers building AI agents that need to perform complex tasks across multiple desktop applications, researchers focusing on computer-use AI, and teams needing isolated, scalable cloud environments for agentic automation.

Highlights

  • Cross-platform support: Works across macOS, Windows, and Linux.
  • Isolated environments: Offers both local VMs (via Lume) and cloud-based sandboxes.
  • Specialized models: Includes CUA-S1 for efficient interface-specific decision making.
  • Agent-agnostic: Allows users to bring their own agent and model while providing the automation infrastructure.
  • Evaluation tools: Built-in benchmarking and RL environments via Cua Bench.

Related

  • Dispatch
    Cua and CUA-S1: Open-Source Infrastructure for AI Computer Use

    Cua is an open-source framework providing desktop automation drivers, isolated cloud and local VMs, and CUA-S1 specialized decision models to enable AI agents to operate computers across macOS, Windows, and Linux.

  • Dispatch
    Driving macOS Apps in the Background: Cua Driver's Solution for Agent-Based Automation

    Cua Driver addresses the long-standing challenge of running UI automation on macOS without disrupting the user's session, enabling agents to interact with applications in the background by leveraging a novel approach to event posting and window management. This innovation paves the way for more seamless and concurrent agent-driven workflows.

  • Project
    Paca-AI/paca

    Paca is an open‑source, self‑hosted project‑management platform that integrates AI agents as first‑class Scrum teammates. It provides a shared Scrumban board, in‑app AI chat, BDD collaboration, role‑based IAM, WASM sandboxed plugins, and an MCP server that lets any compatible AI agent read/write workspace data. Installable via a single Docker‑Compose script or a Helm chart, it’s designed for teams that want AI deeply embedded in their workflow rather than added as a peripheral bot.

  • Project
    jaylfc/taOS

    taOS is a self-hosted AI agent platform that turns spare hardware (laptops, Pis, gaming PCs, phones) into a private distributed compute cluster. It offers a full web desktop environment, framework-agnostic agents, a self-hosted memory system (taOSmd), and large app/model catalogs—all offline-first with cloud as an opt-in. It's beta software with some rough edges and untested install manifests.

  • Project
    openocta/openocta

    A desktop AIOps agent for Windows and macOS that automates IT operations tasks like alert analysis and system inspection through natural language and toolchain integration.