A research prototype of a human-centered web agent

Agent S: an open agentic framework that uses computers like a human

Bytebot is a self-hosted AI desktop agent that automates computer tasks through natural language commands, operating within a containerized Linux desktop environment.

OpenCUA: Open Foundations for Computer-Use Agents

Official implementation of "SEAgent: Self-Evolving Computer Use Agent with Autonomous Learning from Experience"

Code, benchmark and environment for "ScienceBoard: Evaluating Multimodal Autonomous Agents in Realistic Scientific Workflows"

This is the official website for TuriX Computer-use-Agent

Release of code, datasets and model for our work TongUI: Building Generalized GUI Agents by Learning from Multimodal Web Tutorials

OS-Harm: A Benchmark for Measuring Safety of Computer Use Agents

A powerful automation agent for macOS that enables natural language control of various system applications and services. This agent allows you to interact with your Mac using simple text commands, automating tasks across multiple applications including Finder, TextEdit, Preview, and more.

An LLM computer-using agent (CUA) designed to autonomously perform mundane tasks related to business operations and administration, such as doing accounting, filing paperwork, and submitting applications. The accountant is not your bro, but Bro is.

MCP Server for Wayland

🤖 "sudo rm -rf agentic_security" – Investigating computer-use agent security

Computer use Docker Player (LLM Research)

A general-purpose, human-in-the-loop AI-assisted screen interaction toolkit.

Browser automation for YC cofounder matching using OpenAI Computer Use API and Playwright.

下一代MAS(Multi-Agent System)框架，支持MCP,支持无限拓展Agent……详见仓库。

About Mini app where Claude moves the mouse to interact with an HTML page, and uses that interaction to trigger or reflect something in a Flask backend.

Innovative new code that leverages Agents SDK and the computer-use-preview openai api model . The user input a query and the app builds the config and JSON to "visually" search the web for products based on the LLM prompt generated. Test code for proof of concept

Linux GUI for initiating and monitoring Browser Use with an exit switch