# UI-TARS

Pioneering Automated GUI Interaction with Native Agents

Repository: https://github.com/bytedance/UI-TARS
Canonical: https://ross.abutalabs.com/products/ui-tars
Language: Python
License: Apache-2.0
License Family: permissive
Topics: research
Last push: 2026-01-27T04:57:26+00:00

## Health v2 (maintenance only)
Score: 49/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 64, release rhythm 35, longevity 42
- inputs: {"age_days": 591, "days_push": 218, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 11389, forks 870 (observed 2026-08-28T04:10:46.912384+00:00)

## What it is
UI-TARS is ByteDance's open-source multimodal AI agent stack, comprising Agent TARS (a CLI/Web UI multimodal agent that operates terminals, computers, and browsers) and UI-TARS-desktop (a native GUI agent desktop app built on the UI-TARS vision-language model). It connects cutting-edge multimodal LLMs with agent infrastructure and MCP tool integration for human-like task completion.

## Use cases
- automate computer tasks with an AI agent
- control my browser with a vision language model
- run a GUI agent that clicks and types like a human
- use MCP tools with a multimodal AI agent
- operate my desktop remotely with an AI agent
- automate repetitive GUI workflows
- build agents that see and interact with screens

## When to choose
- you want an open-source computer-use or GUI agent driven by vision-language models
- you need MCP tool integration in a multimodal agent workflow
- you want both a CLI and desktop app for AI-driven browser/computer automation
- you want to experiment with the UI-TARS model for screen understanding

## When to avoid
- you need a simple headless API automation without screen interaction
- you require a fully production-hardened enterprise RPA solution with vendor support
- your tasks are purely text-based and don't benefit from vision-based GUI control

## Facets
- artifact type: application
- maturity: active
- function: agent-framework, mcp, computer-vision, llm-inference, gui, cli, web-scraping
- domain: artificial-intelligence, large-language-models, computer-vision, developer-tools, cross-platform
- platform: cross-platform, cli
- tags: computer-use, gui-agent, multimodal, vlm, browser-automation, mcp-server, ui-tars, agent-tars, vision-language-model, desktop-automation, ai-agents, automation, desktop, web-server, nodejs, typescript

## Member repositories
- bytedance/UI-TARS (main) score 49
- bytedance/UI-TARS-desktop (frontend) score 71

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:10:46.912384+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-29T16:24:49.983976+00:00, confidence not recorded.
  - readme: https://github.com/bytedance/UI-TARS (fetched 2026-08-28T04:10:46.912384+00:00, sha e2895e3e92c5)
- Data as of 2026-08-30T08:39:29.467469+00:00.