UI-TARS Desktop

What is UI-TARS Desktop?

UI-TARS Desktop is part of TARS, an open-source multimodal AI agent stack from ByteDance that currently ships two projects. Agent TARS is a general agent that reads web pages visually, calls tools through Model Context Protocol, and works with the command line and file system, so it can plan and carry out jobs such as researching a topic and writing a report. UI-TARS Desktop is the separate desktop application that acts as a GUI agent, controlling a computer through the UI-TARS vision-language model. Agent TARS offers three browser modes, dom, visual-grounding and hybrid, so it can run with or without a vision model, and you can step into a running task with human-in-the-loop input and share the thread as an HTML file. You supply your own model and search provider, including Claude, Azure OpenAI, Tavily or local browser search. The projects are free and open source under Apache-2.0, with no paid plans, and the project's own docs call parts of the stack a technical preview.

Use cases

• Automate desktop tasks with a GUI agent
• Run browser research and write a report
• Operate web pages by visual grounding
• Connect MCP tools to an agent
• Share an agent run as an HTML replay
• Test computer-use models locally

Who it's for

• Developers
• AI researchers
• Automation engineers
• Open-source contributors

Why use UI-TARS Desktop

• Open-source code means you can inspect, modify and self-run the agent without a subscription.
• A visual grounding mode lets the agent click and type from screenshots instead of depending on page markup.
• A dom mode works with models that cannot see images, widening which providers you can use.
• MCP support connects the agent to search, file and command-line tools.
• Shareable replays make it easy to review or demonstrate what the agent did.

UI-TARS Desktop features

• Desktop GUI agent driven by the UI-TARS model
• Agent TARS multimodal agent with CLI and Web UI
• Three browser modes: dom, visual-grounding and hybrid
• MCP integration for search, file editing and CLI tools
• Human-in-the-loop input while a task is running
• Real-time view of browser and document artifacts
• Choice of model provider, including Claude and Azure OpenAI
• Selectable search provider such as Tavily, Bing or SearXNG
• Share a task as local HTML or a remote server URL

UI-TARS Desktop integrations

• Claude
• OpenAI
• Azure OpenAI
• DeepSeek
• Doubao 1.5 VL
• Tavily
• Bing Search
• SearXNG
• DuckDuckGo
• Chrome
• Browser Use

How to get started with UI-TARS Desktop

1. Download the desktop app from the GitHub releases page, or install the Agent TARS CLI.
2. Open Settings and choose a model provider, enter its API key and test the connection.
3. Pick a search provider and a browser control mode: dom, visual-grounding or hybrid.
4. Type a task, such as researching a topic, and watch the agent plan, browse and write results.
5. Step in with human-in-the-loop messages if needed, then share the thread as an HTML file.

UI-TARS Desktop limitations

• The Agent TARS app was described as a technical preview and not recommended for production
• The 2025 docs said the Agent TARS app supported macOS only, with Windows planned
• Agent TARS was best tested with Claude, and other models were experimental
• UI-TARS Desktop works only with the UI-TARS model
• It acts on your computer, so tasks should be supervised

UI-TARS Desktop pricing

UI-TARS Desktop is open source, so you can use and self-host it for free.

Prices can change and may vary by country or exclude tax. Check the official pricing page.

UI-TARS Desktop alternatives

The main UI-TARS Desktop alternatives listed on Debuly are Browser Use, OpenHands, BrowserOS, Hexabot, nanobot and PowerSkills. The table compares pricing and platforms; order is A–Z.

1. 

   Browser Use

   Cloud browsers and hosted web agents for AI, plus an MIT-licensed open-source browser agent, billed pay-as-you-go from $0.02 per hour.

   Developer platformsAutomation and agentsCoding
   PaidVisit Browser Use (official site)
2. 

   BrowserOS

   Free, open-source Chromium browser with built-in AI agents and an MCP server, so agents like Claude Code or Codex can browse as you.

   Automation and agentsProductivityDeveloper platforms
   Open sourceVisit BrowserOS (official site)
3. 

   Hexabot

   Self-hosted platform for building AI agents and workflows that run across chat channels, call tools and keep memory, from $0 to $149/month.

   Automation and agentsDeveloper platforms
   FreemiumVisit Hexabot (official site)
4. 

   nanobot

   Open-source, lightweight AI agent you run yourself, with a web UI, MCP tools, memory, automations and chat app channels.

   Automation and agentsDeveloper platformsChat assistants
   Open sourceVisit nanobot (official site)
5. 

   OpenHands

   Open-source AI coding agent platform with a free Agent Canvas, a Teams plan, and self-hosted Enterprise control

   CodingAutomation and agentsDeveloper platforms
   FreemiumVisit OpenHands (official site)
6. PS

   PowerSkills

   PowerShell toolkit that lets AI agents drive Windows apps, Outlook, Edge and the shell through JSON-returning skills.

   Automation and agentsDeveloper platformsProductivity
   Open sourceVisit PowerSkills (official site)

UI-TARS Desktop FAQ

Sources & verification

Facts on this page come from the official sources below, each with the date we read it. Written by Debuly; the vendor did not review this page.

Image credits

• Logo: none collected yet, so a monogram is shown.
• Cover: none collected yet, so a generated tile is shown.

Spot an error? Report an issue with this listing. We also remove a logo or image on the owner's request.