Bringing Computer Use To The Web

Cua's JavaScript and TypeScript library gives web developers control over a remote Linux desktop running in a Cua cloud container: they can capture its screen and send mouse or keyboard actions. The tutorial connects that library to OpenAI's computer-use model to show the complete cycle.

The application sends a screenshot and task to the model, translates the returned instruction into a desktop action, then takes another screenshot so the model can see what changed. Cua supplies the connection to the computer; the model chooses the actions. This makes desktop automation accessible from a web application's existing language while leaving the developer responsible for connecting those pieces. When the tutorial was written in August 2025, it described that OpenAI model as having limited availability, so access could be a prerequisite.