Setup tutorial by Sriram

UI TARS setup guide

Follow the setup tutorial below, then use the final checklist before relying on it in a real workflow.

  • Practical setup
  • Beginner-readable
  • Business workflow focus
Work with Sriram

Turn your business workflow into an AI-powered system.

Send your requirement and get a tailored AI automation blueprint built for your business. No fluff, just what actually works.

Get this built for my business

Setup guide

UI TARS setup guide

Platform Install path Important note and run it Use admin mode only if control fails Mac Recording Use Ollama when you want a local endpoint. The model must support vision input. Plain text models

Windows and Mac setup

  1. Install

Platform Install path Important note

Windows

Download the latest installer from GitHub Releases

and run it Use admin mode only if control fails Mac

Download from Releases or use Homebrew

Enable Accessibility and Screen

Recording

  1. Prepare Windows or Mac
  • Install Chrome Edge or Firefox before using Browser Operator
  • Use one active monitor for the first test run
  • Mac: System Settings > Privacy and Security > Accessibility and Screen Recording
  • Restart the app after changing Mac permissions
  1. Ollama first

Use Ollama when you want a local endpoint. The model must support vision input. Plain text models are not enough for screen control because the agent needs to understand screenshots.

  • Install Ollama
  • Pull a multimodal vision model
  • Open the app then Settings
  • Use the local Ollama endpoint
  • Click Check Model Availability

VLM Base URL: http://localhost:11434/v1

VLM API KEY: ollama

VLM Model Name: your_ollama_model_name

Ollama note

If the model cannot read images the setup may connect but desktop control will fail or behave badly. Start with a simple visual task before trusting it.

  1. Then OpenRouter

Use OpenRouter for the easiest API router path. The app accepts an OpenAI compatible base URL API key and model name.

  • Create an OpenRouter API key
  • Open the app then Settings
  • Enter the values below
  • Click Check Model Availability
  • Run a harmless test task

Language: en

VLM Provider: Hugging Face for UI-TARS-1.5

VLM Base URL: https://openrouter.ai/api/v1

VLM API KEY: your_openrouter_key

VLM Model Name: bytedance/ui-tars-1.5-7b

  1. Final checks

Check Expected result Model Model availability check passes Browser Agent opens the browser and reads visible page content Desktop Agent sees the screen and can move safely through the interface Permissions Mac permissions are enabled. Windows runs normally unless admin is needed

  1. Simple recommendation

Try Ollama first only if you already have a working local vision model. Use OpenRouter when you want the cleaner and more predictable setup. Safety Start with harmless tasks and keep the screen visible. Do not let the agent operate banking passwords payments email sending or private files without direct supervision.

Final checklist

  • The required app, command, or service opens correctly.
  • Credentials are stored outside public files and screenshots.
  • A small test run completes before you use the setup for real work.
  • You know where logs, output files, and error messages are stored.
Sriram Varadarajan Sriram VaradarajanReplies personally

Want this connected to your real business workflows?

The guide helps you set up the tool. The bigger advantage comes from knowing which automations are worth building for your business.

Map my automation