Ross ROSS = Recommend OSS · open-source software intelligence for agents

trycua/cua

Scale computer-use 2.0 with open-source drivers, cross-OS fleets, and benchmarks for training, evaluation, and data generation. observed · 2026-08-28

github.com/trycua/cua · homepage · HTML · MIT (permissive) observed · 2026-08-28

Health v2 · maintenance only

83/100

  • Activity 99
  • Release rhythm 87
  • Longevity 41
How is this computed?

round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10) — computed 2026-09-03. Adoption (stars, forks) is never an input.

  • gap_med: 0
  • age_days: 579
  • days_rel: 11
  • days_push: 7
  • n_releases_24m: 550

Full methodology

Adoption not part of the score

21922 stars · 1508 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-29, confidence not recorded

Cua is an open-source framework for computer-use agents, providing background desktop drivers, disposable GUI sandboxes, cross-OS fleets (Linux, Windows, macOS, Android), and Cua-Bench for verifiable task benchmarks and RL environments. It includes Lume for Apple Silicon macOS virtualization and a Python SDK for running, evaluating, and generating training data from agent workloads.

Use cases

  • run computer-use agents on real desktops in the background
  • spin up disposable GUI sandboxes for AI agents
  • benchmark and evaluate computer-use models on verifiable tasks
  • generate training data from agent desktop activity
  • run macOS VMs on Apple Silicon locally
  • scale agent evals across parallel cloud fleets
  • let a coding agent drive native desktop apps without taking over the cursor
  • run reinforcement learning environments for desktop agents

When to choose

  • you are building or evaluating computer-use agents and need real or isolated desktop environments
  • you need cross-OS (Linux, Windows, macOS, Android) agent infrastructure with one API
  • you want open-source, self-hostable VM and sandbox runtimes including macOS on Apple Silicon
  • you need repeatable, verifiable benchmarks or RL environments for desktop tasks

When to avoid

  • you only need simple browser automation or web scraping without a full desktop
  • you need a managed no-code RPA product rather than a developer framework
  • your agents only interact with APIs and never need GUI or desktop control

Facets

framework · maturity active

agent-framework machine-learning benchmarking simulation workflow-automation sdk cli mcp artificial-intelligence reinforcement-learning developer-tools cross-platform windows python cloud self-hosted cross-platform computer-use computer-use-agents desktop-automation virtualization macos-vms apple-silicon sandboxing rl-environments agent-evaluation data-generation gui-automation fleets ai-agents automation macos linux android docker

10 sources

Member repositories

RepositoryRoleHealth v2
trycua/cuamain83

For agents

markdown · JSON · MCP: product_card(name="trycua/cua")

Data as of 2026-08-30T08:39:29.467469+00:00 · Report a problem