About this project

TARS is an open-source multimodal AI agent stack that currently ships two related projects: Agent TARS and UI-TARS Desktop. Both are licensed under Apache 2.0 and developed by ByteDance. Agent TARS is described as a general multimodal AI agent stack that brings GUI Agent and Vision capabilities into the terminal, computer, browser and product. It ships primarily as a CLI and a Web UI. According to the README, it aims to provide a workflow closer to human-like task completion through multimodal LLMs and integration with real-world MCP tools. Its stated core features include a one-click out-of-the-box CLI supporting both headful Web UI and headless server execution, a hybrid browser agent that can control browsers via GUI Agent, DOM, or a hybrid strategy, a protocol-driven Event Stream for context engineering and Agent UI, and an MCP-based kernel that supports mounting MCP Servers to connect to external tools. Installation examples use npx or a global npm install (Node.js >= 22), and it can be run with different model providers such as Volcengine or Anthropic by specifying provider, model and API key. A v0.3.0 release notes streaming support for multiple tools, runtime settings with timing statistics, an Event Stream Viewer, and support for an AIO agent Sandbox as an isolated execution environment. UI-TARS Desktop is a desktop application providing a native GUI agent for the local computer, driven by the UI-TARS model and Seed-1.5-VL/1.6 series models. It offers local and remote computer operators as well as browser operators. Its listed features include natural language control powered by a vision-language model, screenshot and visual recognition support, precise mouse and keyboard control, cross-platform support (Windows/macOS/Browser), real-time feedback and status display, and fully local processing for privacy. The README links to a paper (arXiv 2501.12326), Hugging Face models, ModelScope, and a Discord community. The README also references a UI TARS SDK as a cross-platform toolkit for building GUI automation agents, and documentation covering quick start, deployment, and MCP integration. Example use cases shown include booking flights or hotels, generating charts, and adjusting application settings through natural-language instructions.