UFO: Microsoft's Multi-Device AI Agent Orchestration Framework

This repository profile is provided by osrepos.com, an open source repository discovery platform.

UFO: Microsoft's Multi-Device AI Agent Orchestration Framework

Summary

Microsoft's UFO project introduces a powerful framework for intelligent automation, evolving from a robust Windows Desktop AgentOS (UFO²) to a revolutionary Multi-Device Agent Galaxy (UFO³). This project enables the orchestration of AI agents across diverse platforms, streamlining complex workflows and enhancing digital interaction. It offers both standalone Windows automation and a scalable solution for cross-device collaboration.

Repository Information

Analyzed by OSRepos on March 20, 2026

Topics

Click on any tag to explore related repositories

Use at your own risk

OSRepos shares public repositories for knowledge and discovery only. Any installation, execution, configuration, or use of code from these repositories is the user's own responsibility. Always review the repository, source code, dependencies, licenses, and security implications before running or installing anything. OSRepos is not responsible for issues, damages, or losses resulting from third-party repositories.

Introduction

UFO, short for "UFO³: Weaving the Digital Agent Galaxy," is an innovative open-source project from Microsoft designed to revolutionize intelligent automation. Written primarily in Python, UFO has garnered significant attention with over 8,100 stars and 1,000 forks, reflecting its impact in the AI agent and automation landscape.

The project has evolved through three major phases:

  • UFO (February 2024): The original UI-Focused agent for Windows.
  • UFO² (April 2025): Evolved into a Desktop AgentOS, offering stable and battle-tested single Windows automation.
  • UFO³ Galaxy (November 2025): The latest and most advanced iteration, introducing a multi-device orchestration framework capable of coordinating intelligent agents across heterogeneous platforms.

UFO³ combines the power of the new Galaxy framework for multi-device orchestration with the proven capabilities of UFO² as a robust Windows device agent. This allows users to tackle everything from simple desktop tasks to complex, cross-device workflows.

Installation

Getting started with UFO involves choosing between the UFO³ Galaxy for multi-device orchestration or UFO² for Windows-specific automation. Both require Python and LLM API configuration.

UFO³ Galaxy Quick Start (For cross-device orchestration)

  1. Install Dependencies:
    pip install -r requirements.txt
    
  2. Configure ConstellationAgent:
    copy config\galaxy\agent.yaml.template config\galaxy\agent.yaml
    

    Edit config\galaxy\agent.yaml to add your API keys (e.g., OpenAI or Azure OpenAI).

  3. Configure Devices:

    Edit config\galaxy\devices.yaml to register your devices (Windows, Linux, Android).

  4. Start Device Agents:

    Follow platform-specific guides to start server and client components for each device.

  5. Launch Galaxy:
    python -m galaxy --interactive
    

UFO² Quick Start (For Windows automation)

  1. Install Dependencies:
    pip install -r requirements.txt
    
  2. Configure LLMs:
    copy config\ufo\agents.yaml.template config\ufo\agents.yaml
    

    Edit config\ufo\agents.yaml to add your API keys.

  3. Run UFO²:
    python -m ufo --task <task_name>
    

Common LLM Configuration

Both frameworks require LLM API configuration. Here's an example for OpenAI:

For Galaxy (config/galaxy/agent.yaml):

CONSTELLATION_AGENT:
  REASONING_MODEL: false
  API_TYPE: "openai"
  API_BASE: "https://api.openai.com/v1/chat/completions"
  API_KEY: "sk-your-key-here"
  API_MODEL: "gpt-4o"

For UFO² (config/ufo/agents.yaml):

VISUAL_MODE: True
API_TYPE: "openai"
API_BASE: "https://api.openai.com/v1/chat/completions"
API_KEY: "sk-your-key-here"
API_MODEL: "gpt-4o"

More LLM options (Qwen, Gemini, Claude) are available in the official documentation.

Examples

UFO³ Galaxy excels at orchestrating complex workflows across multiple devices, breaking down tasks into executable DAGs (Directed Acyclic Graphs) and coordinating agents on different platforms. For instance, it can manage a task that requires data extraction on a Linux server, processing on a Windows machine, and final reporting on a mobile device.

UFO², on the other hand, is optimized for single Windows automation, performing tasks like interacting with GUI elements, automating application workflows, and integrating with native Windows OS functionalities. It can serve as a powerful device agent within the larger UFO³ Galaxy framework.

You can watch UFO³ Galaxy in action orchestrating cross-device tasks on their official YouTube channel.

Why Use UFO?

UFO offers a versatile solution for intelligent automation, catering to different needs:

UFO³ Multi-Device Agent Galaxy (New & Recommended):

  • Cross-device collaboration: Ideal for workflows spanning multiple operating systems and devices.
  • Complex multi-step automation: Handles intricate tasks with DAG-based orchestration.
  • Heterogeneous platform integration: Supports Windows, Linux, Android, and more.
  • Dynamic DAG editing: Adapts workflows based on execution feedback.
  • Unified AIP protocol: Ensures secure and fault-tolerant agent communication.

UFO² Desktop AgentOS (Stable & Battle-Tested):

  • Single Windows automation: Perfect for desktop-specific tasks.
  • Quick task execution: Streamlined for rapid automation.
  • Deep Windows OS integration: Leverages UIA, Win32, and WinCOM for robust control.
  • Hybrid GUI + API actions: Combines visual interaction with programmatic calls for efficiency.
  • Long-Term Support (LTS): Actively maintained with ongoing bug fixes and improvements.

UFO² can also seamlessly serve as a Windows device agent within the UFO³ Galaxy framework, providing a flexible migration path for users looking to scale their automation capabilities.

Links

Related repositories

Similar repositories that may be relevant next.

Hermes WebUI: A Powerful Web Interface for Your Autonomous AI Agent

Hermes WebUI: A Powerful Web Interface for Your Autonomous AI Agent

June 1, 2026

Hermes WebUI provides a lightweight, dark-themed web application for interacting with Hermes Agent, offering full parity with the CLI experience. This self-hosted interface allows users to manage sessions, browse workspaces, and control their AI agent from any web browser or phone, enhancing accessibility and user experience. It integrates seamlessly with existing Hermes Agent setups, requiring no additional configuration.

agentai-agentshermes
Hexabot: Open-Source AI Chatbot and Agent Builder

Hexabot: Open-Source AI Chatbot and Agent Builder

March 19, 2026

Hexabot is an open-source AI chatbot and agent builder designed for creating and managing multi-channel and multilingual conversational agents with ease. It offers extensive customization, powerful text-to-action capabilities, and supports integration with various LLM models, making it a flexible solution for developers. This project simplifies the deployment and management of sophisticated AI-powered interactions across different platforms.

aichatbotagent
mini-swe-agent: The Minimal AI Agent for Solving GitHub Issues

mini-swe-agent: The Minimal AI Agent for Solving GitHub Issues

March 18, 2026

mini-swe-agent is a remarkably simple yet powerful AI agent, comprising just 100 lines of Python code. It's designed to solve GitHub issues and assist in command-line tasks, achieving over 74% on the SWE-bench verified benchmark. This project offers a radically simple approach to AI-driven software engineering, avoiding complex configurations and large monorepos.

agentai-agentpython
learn-claude-code: Build AI Coding Agents from Scratch with Python

learn-claude-code: Build AI Coding Agents from Scratch with Python

January 24, 2026

The learn-claude-code repository offers a progressive tutorial to demystify AI coding agents like Claude Code, Kode, and Cursor Agent. It teaches users how modern AI agents work by building them from scratch, starting with a minimal 16-line Bash agent. This project emphasizes the core concept of "Model as Agent" through five evolving versions.

agentclaude-codeteaching

Source repository

Open the original repository on GitHub.

View on GitHub
OS
OSRepos

Analysis and discovery of open source repositories. Find interesting projects and follow their updates.

Monitor your website with YourWebsiteScore

OSRepos shares public repositories for knowledge and discovery only. Any installation, execution, configuration, or use of third-party repository code is at your own risk. Always review source code, dependencies, licenses, and security implications before running anything.

© 2025 OSRepos. Built with Nuxt 3 and lots of ❤️