💬UI-TARS
bytedance/UI-TARS-desktop · homepage ↗
The Open-Source Multimodal AI Agent Stack: Connecting Cutting-Edge AI Models and Agent Infra
⭐ Very popular: 39k stars, gaining about 131 a week
View on GitHub ↗repo profile
momentum
durability
bus factor = how many people it takes to cover more than half the commits (6 months). 1 is a solo project; higher means the work is spread across a team. top-author share is the single busiest author's slice of those commits.
since we covered it
why it's a big deal
- Gives developers an open-source stack for building AI agents that operate real computer interfaces, covering both a native GUI-control desktop app and a general Agent TARS framework.
- Targets people automating on-screen tasks through vision-based control of mouse and keyboard, with cross-platform support across Windows, macOS, and browser.
- Ships CLI, Web UI, and headless server modes plus MCP integration, so external tools can be connected without building the agent runtime from scratch.
under the hood
- Written mostly in TypeScript and organised as a pnpm and Turbo monorepo split into apps, packages, multimodal, and infra directories, with Vitest for tests.
- Uses Vision-Language Models to drive computer control from screenshots, and offers a hybrid browser agent that can run via GUI Agent, DOM, or a combined strategy.
- Runs on the UI-TARS-1.5 and 1.6 Seed models and integrates model providers including Volcengine DouBao and Anthropic Claude, with an Event Stream protocol for context handling.
Radar summary, generated from the project's public sources
star history
- PR#1 2k 2025-02-05
- now 39k + 36k since first covered
curve is sampled from GitHub's star history, plus our own daily readings since we covered it; the dashed stretch is before we first covered it, the solid line since. figures at coverage are the numbers we printed then (approx.), current count is live.
understory
Better known than its recent output, coasting a little on attention.
- output, commits & releases
- clout, star velocity
output = commits & releases; clout = star velocity, both 0 to 100 monthly indices; the gap where output runs above clout is the understory. The understory →
covered in
-
Control Your Computer With Natural Language
similar projects
compare these →- 📖 ai-agent-book
Python
AI Agent PDF
38k ACTIVE - 🌐 worldmonitor
2.1× the stars · AGPL-3.0 licensed
Real-time global intelligence dashboard. AI-powered news aggregation, geopolitical monitoring, and infrastructure tracking in a unified situational awareness interface
82k ACTIVE - 🧠 Context Mode
leaner, 20k stars
Context window optimization for AI coding agents. Sandboxes tool output (98% reduction), persists session memory, and enforces routing across 17 platforms via MCP + hooks.
20k ACTIVE
comments
Sign in with GitHub to add your blip on UI-TARS.