UFO review
A flexible open-source framework for Windows, Android, and multi-device automation.
Reviewed by iTechGuides Editors · Editorial team · Updated Oct 2026
UFO is Microsoft’s open-source framework for natural-language computer-use automation. It is aimed at developers and researchers who want to run and customize agents locally or across connected devices. UFO² focuses on Windows desktop workflows across multiple applications, while UFO³ extends orchestration across Windows, Linux, and Android. Its platform coverage includes desktop app control, browser control, computer vision, API access, and human approval for sensitive actions.
The framework’s standout capability is its combination of visual and native execution. Agents can capture screenshots, detect UI controls, use Windows UI Automation, Win32, and WinCOM, and switch between GUI actions and native APIs. Multimodal visual reasoning is configurable, while retrieval-augmented generation can draw on help documents, web search, demonstrations, and execution traces. For broader environments, HTTP dispatch connects tasks to devices, and multi-device graphs coordinate agents through task graphs and AIP. Android automation is supported through ADB-based MCP servers.
UFO is open-source rather than a hosted subscription product. Users configure their own LLM provider, with integrations including MCP, OpenAI, Azure OpenAI, Gemini, Claude, and Qwen; separate provider API costs may apply. That model suits teams that want local control, customization, and an extensible automation framework. It is less suitable for buyers seeking a managed product with a preconfigured model service and minimal setup. Choose UFO when Windows automation, Android coordination, native API access, or multi-device workflows are central requirements; consider a narrower tool when those capabilities would add unnecessary configuration.
UFO pros and cons
- Where it wins
- Combines GUI controls, native APIs, screenshots, and visual reasoning
- Coordinates Windows, Linux, and Android agents through task graphs
- Supports MCP and multiple model providers, including OpenAI and Claude
- Where it doesn't
- Requires users to configure their own LLM provider
- Separate model-provider API costs may apply
- Developer and researcher focus can mean more configuration work
UFO fact sheet, pricing and score →
Advertiser disclosure: iTechGuides is reader-supported. Vendors can pay for top positions in our rankings and for a place on other products' pages, and we may earn a commission when you click some links. Bought places stand at the number bought; this page and our disclosure say so. Paying vendors may get a fuller, more detailed review from our editors, and that review can change their score. How we rank.
Last updated · How we research and update