computer-using-agent 开源项目推荐（14 个）

a-real-ai /pywinassistant

第一个开源的人工狭义智能通用代理框架Computer-Using-Agent，仅使用自然语言完全操作图形用户界面（GUI）。使用思想可视化和思维链推理来引发空间推理和感知，模拟，计划和模拟合成HID交互。

OS-Agent-Survey /OS-Agent-Survey

这是论文《OS Agents：基于 MLLM 的计算机、手机和浏览器使用代理综述》（ACL 2025 Oral）的仓库。

OS-Copilot /ScienceBoard

[ICLR 2026]《ScienceBoard：在现实科学工作流程中评估多模态自主智能体》的代码、基准与环境

fyzhang1 /DualTAP

[ECCV 2026] Official Repository for "DualTAP: A Dual-Task Adversarial Protector for Mobile MLLM Agents"

fyzhang1 /ExpActivator

Official code for "Relevance Does Not Imply Applicability: Experience Activation for Personal GUI Agents".

reidbarber /webmarker

Mark web pages for use with vision-language models

OS-Copilot /OS-Symphony

[ACL 2026 Main] Official repository for paper: OS-Symphony: A Holistic Framework for Robust and Generalist Computer-Using Agents

cocoa-org /NanoRollout

Scale digital agent rollouts without pain.

ahnjaewoo /FlashAdventure

🕵 Code for our EMNLP 2025 Main paper: "FlashAdventure: A Benchmark for GUI Agents Solving Full Story Arcs in Diverse Adventure Games"

Clad3815 /open-computer-use

AI-powered assistant that controls a Windows environment through docker, allowing automated interaction with the desktop interface. Control your computer with natural language.

ercbot /valk

Simple, observable computer use - Remote desktop for AI agents

Mihonarium /food_ordering_agent

使用LLM代理自动订购Deliveroo、Uber Eats、DoorDash等的食品和其他商品。

1azybug /dsh-real-time-computer-use

A plugin used for the real-device evaluation in “On the Capability Boundaries of Turn-Based GUI Agents in Real-Time GUI Environments: A Human Perception–Action Perspective”

thethinkmachine /4o-agent

A ReAct based CUA built with LangChain