whereismyai
Cua
Cua is an open-source computer-use driver that scales Linux, Windows, macOS, and Android fleets for AI agents.
Cua scales Linux, Windows, macOS, and Android computer fleets for computer-use agents using an open-source MCP/CLI driver. It runs workloads in parallel, forks from snapshots, reproduces failures, and turns agent activity into training data. The product supports Ubuntu containers and VMs, Windows 11 and Server 2025, Apple Silicon virtualisation, and Android VMs. It integrates with tools like OpenAI, Qwen Code, and Factory Droid. Cua provides open-source control and eval layers including Cua Driver, Cua Bench, and Lume, operating via local sandboxes or the Cua Fleet cloud service.
Cua at a glance
- Best for: Software engineer, Researcher
- Runs on: macOS, Windows, Linux, Android, API, CLI, self-hosted, open source
- Pricing: $0.044625/vCPU/hour; $0.0223125/GiB/hour
- Released: January 2025
- Category: AI agent tool
- Website: cua.ai
Key features
- Scale fleets across Linux, Windows, macOS, and Android
- Run Ubuntu containers or full VMs
- Run Windows 11 and Server 2025 for native apps
- Run Sequoia and Tahoe on Apple Silicon virtualization
- Run Android VMs and automate touch workflows
- Inspect accessibility trees and capture window state
- Operate as an MCP stdio server or shell command
Use it for
- Run computer-use agent training and evaluation
- Generate training data from agent activity
- Automate browser and code work
- Execute native desktop app workflows
- Automate Android taps, swipes, and multi-touch
Alternatives to Cua
- AutoCua — It uses operating system input layers and accessibility trees instead of pixel guessing.
- NemoClaw — Provides managed security, networking, and lifecycle operations for AI agents.
- OpenGCUA — Lets models decide how to observe, act, and organize work dynamically.
- AionUi — Auto-detects and runs local CLI agents like Claude Code and Codex in parallel.
- BrowserOS — Runs entirely locally and lets AI agents use real user logins to complete tasks.
- AgentDock — It coordinates multi-device workflows and provides secure environment access.
- Mac Computer Use — Enables local desktop automation through native Mac windows and Accessibility controls.
- OpenViking — Unifies agent memory, knowledge RAG, and skills into a single directory structure.
- Agenta — Build and automate AI agents that run on your own infrastructure or the cloud.
- Agensh — Scales to 1,024 self-organizing agents without a central orchestrator.
- Composio — Equips AI agents with tools, managed user auth, and sandboxes across 1,500+ apps
- FastGPT — It uses hybrid retrieval and auto-cleansed data to prevent hallucinations.
All Cua alternatives