Open-source infrastructure for Computer-Use Agents. Sandboxes, SDKs, and benchmarks to train and evaluate AI agents that can control full desktops (macOS, Linux, Windows).

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.

MagenticLite is an experimental agent that works across the browser and local file system

Fara-7B: An Efficient Agentic Model for Computer Use

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

👾 Open Computer Use – Open-Source Alternative to Codex Computer Use

The first open-source Artificial Narrow Intelligence generalist agentic framework Computer-Using-Agent that fully operates graphical-user-interfaces (GUIs) by using only natural language. Uses Visualization-of-Thought and Chain-of-Thought reasoning to elicit spatial reasoning and perception, emulates, plans and simulates synthetic HID interactions.

Surf is a computer use AI agent powered by OpenAI that interacts with a E2B's virtual desktop environment through natural language instructions

Enable AI to control your desktop, mobile and HMI devices

Stop AI agents from doing things you didn't ask for.

OSWorld 2.0: Benchmarking Computer Use Agents on Long-Horizon Real-World Tasks

[NeurIPS 2025] Official repository of RiOSWorld: Benchmarking the Risk of Multimodal Computer-Use Agents

[NeurIPS'25 D&B] Mind2Web-2 Benchmark: Evaluating Agentic Search with Agent-as-a-Judge

ncurses (pdcurses) based library for CUA look menu

OS-Harm: A Benchmark for Measuring Safety of Computer Use Agents [NeurIPS 2025 Spotlight]

Emacs mode that moves to modern keybindings

Mark web pages for use with vision-language models

A lightweight Python library to create, run and manage macOS and Linux virtual machines (VMs) natively on Apple Silicon.

Opensource benchmark evaluating web operators/agents performance