Cua
Scale computer fleets for AI agents across Linux, Windows, macOS, and Android
Visit cua.ai ↗What is Cua?
Cua provides infrastructure for running computer-use agents at scale. It includes Cua Driver (open-source background automation), Cua Sandbox (disposable GUI environments), and Cua Fleets (elastic pools of pre-booted machines). Users can run agent training, evaluation, and data generation workloads in parallel across multiple operating systems through a single API, with local runtimes or cloud deployment options.
Built for AI teams and researchers working with computer-use agents. Cua Bench evaluates agent performance on tasks like KiCad workflows. The platform supports Linux containers and VMs, Windows 11 and Server 2025, macOS on Apple Silicon, and Android VMs with touch automation. Open-source control and evaluation layers; MIT licensed with on-premises and SOC 2 Type I BYOC options available.
Momentum
Daily activity over the last 30 days: upvotes, unique visits, new GitHub stars and Hacker News points.
How do we know it's vibe coded?
vibetrend discovered Cua automatically. Evidence: Trending on gittrend.io: agent files (claude-code) in trycua/cua. View source.
Similar vibe-coded apps
Stateful AI agents with persistent memory that learn and improve over time
Enterprise AI agent platform built in Go with multi-tenant security
Self-hosted AI model registry for enterprise inference with vLLM and SGLang
Operating system for building and running AI agents across 500+ models
Open-source platform for building and deploying AI-powered web applications