█

AI/TLDR

UI-TARS Desktop

An open-source GUI agent stack that controls your computer and browser with vision-language models

github.com/bytedance/UI-TARS-desktop★ 38kagent-tars.com

Overview

UI-TARS Desktop is a desktop application from ByteDance that gives you a GUI agent: software that looks at the screen and operates a computer or browser the way a person would, using mouse and keyboard actions. It is built on the UI-TARS vision-language model, so the agent works from what it sees on screen rather than from a fixed API.

It is part of a wider TARS stack. Alongside the desktop app sits Agent TARS, a multimodal agent that runs in the terminal and browser through a CLI and Web UI, with support for MCP tools. The desktop app can drive a local computer and browser, or connect to a free remote computer and browser operator with no setup.

It fits the computer-and-browser-use corner of the agent-frameworks space. It is aimed at developers and researchers who want to automate on-screen tasks, test GUI agents, or build their own automation flows on top of a vision-driven model.

What it does

  • Native GUI agent built on the UI-TARS vision-language model, driving a computer through on-screen actions
  • Local operators for controlling your own computer and browser
  • Remote computer and browser operators that need no configuration and are free to use
  • Agent TARS companion stack with a CLI and Web UI for terminal and browser tasks
  • Works with multiple model providers, including Volcengine and Anthropic, via Agent TARS
  • MCP tool integration in Agent TARS for connecting real-world tools

Getting started

The TARS stack ships two parts. The Agent TARS CLI installs from npm and is the quickest way to start; the UI-TARS Desktop app is installed from its own quick-start guide.

Try Agent TARS with npx

Run the CLI directly without installing it. Requires Node.js 22 or newer.

bashbash
npx @agent-tars/cli@latest

Install the CLI globally

Install once to get the agent-tars command on your PATH.

bashbash
npm install @agent-tars/cli@latest -g

Run with a model provider

Start Agent TARS pointing at a model provider and your API key. The README shows Volcengine and Anthropic examples.

bashbash
agent-tars --provider anthropic --model claude-3-7-sonnet-latest --apiKey your-api-key

Install the UI-TARS Desktop app

For the native desktop GUI agent with local and remote operators, follow the project's Quick Start guide at docs/quick-start.md in the repository.

Commands and code are distilled from the project's own documentation — always check the official repo for the latest.

When to use it

  • Automating repetitive on-screen tasks across desktop apps that have no API
  • Running browser workflows that need a model to read and click real page elements
  • Researching and testing GUI agents driven by vision-language models
  • Trying remote computer or browser control without setting up a local environment

How UI-TARS Desktop compares

UI-TARS Desktop alongside other open-source computer & browser use tools AI/TLDR tracks, ranked by GitHub stars.

ToolStarsWhat it does
Browser Use★ 105kA Python library that lets agents control a real browser to read pages and complete tasks online from natural-language instructions.
Open Interpreter★ 65.9kA lightweight coding agent that runs code on your own computer to carry out tasks from natural-language requests.
UI-TARS Desktop★ 38kAn open-source GUI agent stack that controls your computer and browser with vision-language models
Stagehand★ 23.5kA TypeScript browser-automation SDK from Browserbase that mixes natural-language actions with normal code for reliable web agents.
Skyvern★ 22.4kA tool that uses language models and computer vision to automate browser workflows without writing custom code for each website.
Cua★ 19.9kInfrastructure for computer-use agents, providing sandboxes and SDKs that let agents control full desktops on macOS, Linux, and Windows.
Agent Zero★ 18.4kA general-purpose personal agent framework that uses the computer, terminal, and code as its main tools to complete tasks.
Browser Use Web UI★ 16.2kA user-friendly web UI, built on Gradio, that lets you control browser-use AI agents, pick from many LLMs, and use your own browser.