Agent TARS
VerifiedOfficialby ByteDance
Official listing: built and published by ByteDance. More from ByteDance
ByteDance's multimodal AI agent for desktop and browser automation — controls GUIs, browsers, and shell via vision and MCP tools. Available as npm CLI and desktop app.
- Access
- CLI
- Auth
- None
- Skills
- 4
Skills
GUI Control
Control desktop application interfaces using vision-based element detection, clicking, and keyboard input.
Browser Agent
Navigate websites, fill forms, click elements, and extract data from web pages autonomously via the browser.
Shell Execution
Execute shell commands and scripts as part of multi-step agent workflows with streamed output support.
MCP Tool Integration
Connect to local or remote MCP servers and invoke their tools as steps in multi-tool agent pipelines.
Related Agents
UI-TARS Desktop
by ByteDance
ByteDance's multimodal GUI agent stack that controls desktop and browser interfaces from natural language using…
Agent S
by Simular AI
Operates desktop graphical interfaces like a person, combining hierarchical planning with visual grounding to click…
Microsoft UFO
by Microsoft
Automates Windows desktop apps through natural-language tasks with the UFO2 Desktop AgentOS, and orchestrates agents…
Ghost OS
by Ghostwright
Lets any MCP-capable agent see and operate every macOS app through the accessibility tree, with vision fallback and…