AI Agents
MobileAgent
Mobile-Agent — multimodal GUI agents that operate phones and PCs from screenshots and instructions.
★ 9.3k starsMIT
View guide →Microsoft's GUI agent that grew into UFO³ — orchestrating agents across Windows, Linux and mobile devices.
UFO is an open-source ai agents project by microsoft (microsoft/UFO on GitHub). It has 9,961 GitHub stars and 1,068 forks, is released under the MIT license, and was last updated on GitHub on 7 October 2026.
Decide whether you need single-device Windows automation (UFO²) or cross-device orchestration (UFO³ Galaxy), and which devices and LLM you will connect.
Clone microsoft/UFO from GitHub (or use the install path in the run steps below) and skim its README for the parts relevant to your goal.
Clone the repo, install the Python requirements, then copy the Galaxy agent template and add your API keys, and register devices in the devices config.
Start the device agent server and client on each device (with the platform flags for Windows, Linux or Android), then give the ConstellationAgent a task and watch it split into a DAG.
The Galaxy framework coordinates agents across multiple devices as one system, rather than automating a single desktop in isolation.
Tasks are decomposed into executable DAGs, with dynamic DAG editing so the workflow can adapt as execution evolves.
Constellation runs tasks asynchronously with parallel coordination across device agents.
Agents communicate over a unified AIP protocol designed for secure agent-to-agent communication.
The stable single-device path offers deep Windows OS integration and hybrid GUI plus API actions for reliable desktop automation.
Device agents run on Windows, Linux and mobile (Android), each starting a server, MCP servers where applicable, and a client.
Get the source from the official GitHub repository for UFO.
git clone https://github.com/microsoft/UFO.git
cd UFOInstall dependencies, copy the Galaxy agent config template and add your API keys, as the README's quick start shows.
pip install -r requirements.txt
copy config\galaxy\agent.yaml.template config\galaxy\agent.yaml # then edit in your API keysEdit the Galaxy devices file to register your devices, then start each device agent with its platform flags per the platform guides in the docs.
# Edit config\galaxy\devices.yaml to register your devices
# Then start server + client on each device — see README platform guidesMicrosoft, through the microsoft/UFO repository, with full documentation at microsoft.github.io/UFO.
Yes — the project is open source under the MIT license as recorded for this listing. You can use, modify and distribute it under the MIT terms.
You need your own LLM API keys, added to the Galaxy agent configuration — UFO itself is free, but model usage is billed by your provider.
Single-device agents automate one desktop; UFO³ adds the Galaxy layer — task DAGs, parallel coordination and device agents across Windows, Linux and Android — while UFO² remains its proven single-Windows-device mode.
Clone the repository (or use the documented installer), follow the install and configure steps on this page, and run the smallest real task first — start the device agent server and client on each device (with the platform flags for Windows, Linux or Android), then give the ConstellationAgent a task and watch it split into a DAG.
Mobile-Agent — multimodal GUI agents that operate phones and PCs from screenshots and instructions.
Open-source Computer Use as an MCP service: let any agent drive macOS, Linux, and Windows desktops.
Android GUI agent framework: AI that sees, plans, and taps real mobile apps.
ByteDance's open-source multimodal AI agent stack: GUI agents that see and operate real desktops
Control Android and iOS devices with natural language — an LLM-agnostic mobile agent framework and CLI.
Microsoft's trainer for AI agents - reinforcement learning for almost any agent framework
The most-starred other AI Agents projects in the LibreAI catalogue — the realistic shortlist if UFO is not the right fit.
A personal AI that really does things — on any OS, with your data staying yours.
An agentic skills framework & software development methodology that works.
Agent harness optimization system with skills, memory and security for Claude Code, Codex and more
The agent that grows with you — a Python framework for adaptive AI agents.
DeepSeek's plugin-based agent harness where everything is a plugin.
Free and open — grab the source, star the repo, and start creating.