Share this page!

Enter URL or ID to Unroll

You can paste full URL like: https://x.com/threadreaderapp/status/1644127596119195649
or just the ID like: 1644127596119195649

How to get URL link on X (Twitter) App

  1. On the Twitter thread, click on or icon on the bottom
  2. Click again on or Share Via icon
  3. Click on Copy Link to Tweet
  4. Paste it above and click "Unroll Thread"!
  5. More info at Twitter Help
Cua Profile picture
Cua
@trycua
Open-source infrastructure for Computer-Use Agents // YC X25
May 27 • 7 tweets • 3 min read
Today we're bringing Cua Driver to Windows: background computer-use for any agent. Claude Code, Codex, or your own loop can drive real Windows apps through CLI or MCP while your desktop stays usable, with true multi synthetic pointer support.

1/6Image Windows has a lot of Windows inside it. Win32, WPF, WinUI, UWP/WinRT, Electron, Chromium, legacy controls, custom-rendered canvases.

A bunch of us at Cua are ex-Microsoft engineers, and Windows was still harder to tackle than macOS.

Plug Cua Driver into a coding agent or general agent, and the model gets a much wider loop to think with: code, pixels, accessibility trees, app state, clicks, typing, verification.

This is what it looks like:

2/6
Apr 23 • 8 tweets • 3 min read
We're open-sourcing Cua Driver - our new macOS driver that lets any agent (Claude Code, Codex, your own loop) drive any app in the background, with true multi-player and multi-cursor built-in.

1/8 Image Codex introduced background computer-use last week. It's OpenAI-only. We got the same primitive working open-source via a mix of SkyLight private APIs and yabai's focus-without-raise pattern. Kudos to the Sky team (@AriX et al) for originally proving loud first that a UX for background computer-use was possible.

2/8
Dec 16, 2025 • 6 tweets • 3 min read
Today we're announcing cua-bench: a framework for benchmarking, training data, and RL environments for computer-use AI agents. Why? Current agents show 10x variance across minor UI changes. Here's how we're fixing it. 1/ The problem: Claude Computer-Use, OpenAI CUA, and Gemini 2.5 Computer-Use show 10x performance difference across minor UI changes.

Completes tasks on a clean desktop, fails when windows overlap. Works on default themes, breaks on high contrast mode.

The root cause: training data lacks visual diversity.Image
Jun 1, 2025 • 6 tweets • 2 min read
Introducing App-Use - create virtual desktops for AI agents to focus on specific apps.

1/6 App-Use lets you scope agents to just the apps they need. Instead of full desktop access, say "only work with Safari and Notes" or "just control iPhone Mirroring" - visual isolation without new processes for perfectly focused automation.

2/6
May 29, 2025 • 6 tweets • 2 min read
Launching public preview of C/ua Cloud Containers today - Docker for Computer-Use Agents. Zero local setup. Same Computer and Agent interfaces. Scale 1-100 agents instantly.

1/6 Image First cloud platform built for Computer-Use Agents. Open-source backbone. Linux/Windows/macOS desktops in your browser. Works with OpenAI, Anthropic, or any LLM. Pay only for compute time.

2/6
May 2, 2025 • 5 tweets • 1 min read
Want to help make AI better at using computers? We just released a guide on creating human trajectory datasets with C/ua Image Our new @Gradio UI makes it easy to record, organize and share high-quality trajectories of how humans navigate interfaces, recover from errors, and complete real tasks.
Apr 10, 2025 • 7 tweets • 1 min read
Cua loves MCP. Launching official MCP Server support today - enabling our Computer-Use Agent to run through Claude Desktop, Cursor, and other MCP clients. An example use case? How about using Claude as a tutor to learn how to use Tableau?