GitHub Trending · Agent Tooling · Updated
trycua/cua
Give AI agents computers they can use: open-source drivers, cross-OS fleets, and benchmarks
- Stars
- ★ 28.9k
- Language
- Rust
- Chart appearances
- 4
- On-chart tracking
- 1 days tracked
What it does
Cua is an open-source infrastructure stack for computer-use agents, letting AI agents operate desktop applications like a human. It ships desktop automation drivers, isolated cloud desktop fleets, local macOS VMs, specialist decision models, and benchmarks spanning training, evaluation, and data generation.
What's inside
The repo is organized into modules: Cua Driver (cross-platform app operation on macOS/Windows/Linux), Cua Fleets (isolated cloud desktops), Lume (local macOS/Linux VMs on Apple Silicon), the CUA-S1 small model, and Cua Bench for evaluation, plus docs and examples.
Tech stack
HTML-heavy presentation with Python/TypeScript SDKs at the core, relying on virtualization (Lume on Apple Silicon), containerized cloud desktops, and multimodal model inference; offers Docker and cloud deployment paths.
Use cases
- AI agent developers: give agents real desktop control plus an evaluation loop
- Researchers: generate trajectory data and benchmark computer-use models with Cua Bench
- Enterprise teams: run automation and regression tests at scale on isolated cloud desktops
- Mac users: run local macOS/Linux VMs on Apple Silicon with Lume
Why it's trending
With 1,211 stars added this week, a total past 27k, and a #15 spot on the GitHub Trending weekly chart, it shows computer-use agents moving from demos to scalable infrastructure and strong demand for open-source alternatives.
Current chart placements
| Period | Language filter | Rank | Stars gained |
|---|---|---|---|
| Daily | All languages | #10 | +228 |
| Daily | Rust | #1 | +228 |
| Monthly | All languages | #19 | +6.4k |
| Monthly | Rust | #2 | +6.4k |
Editorial summaries are based on the public repository and the dated GitHub Trending snapshot. Verify implementation details in the upstream repository.