Skip to content

UI-TARS-desktop

Control a computer with natural language through a vision-language model

Visit official site

What UI-TARS-desktop is

The app is described as a desktop client developed by ByteDance and built on the UI-TARS vision-language model, letting users interact with a computer through natural language. The app understands instructions and uses screenshots and visual recognition to carry out precise mouse and keyboard actions.

What you can do with it

  • Control computer actions with natural language instructions, such as opening an app or entering text
  • Use screenshots and visual recognition to identify on-screen elements and interact with them
  • Apply precise mouse and keyboard control that simulates a real user
  • Check real-time feedback on the status of operations

Application scenarios

  • Open a browser with a natural language instruction and search for specific content
  • Post a message on social media through a voice instruction
  • Fill in a form automatically and submit the data

Getting started

  • Download the latest UI-TARS-desktop installer from GitHub
  • Install it according to the operating system, Windows or macOS
  • Grant permissions such as screen recording and accessibility after launch
  • Open the app and enter a natural language instruction

Pros & cons

✓ What we like

  • Precise mouse and keyboard control simulates real user actions
  • Real-time feedback shows the status of each operation

! What to watch out for

  • The client is built on the UI-TARS vision-language model
  • Operation relies on screenshot capture and visual recognition

FAQ

Which operating systems are supported?

The material states cross-platform support for Windows and macOS.

How do I get the app?

The documented steps start with downloading the latest installer from GitHub, where the project is open source and free to download.

Last reviewed: 2026-11-15

More AI office assistant tools

View all →

How we review