Lets an AI vision model operate a real iPhone using a $7 ESP32-S3 board that acts as an AirPlay screen receiver and USB HID mouse, with no computer needed at run time.
chekusu/yomado
A native macOS app that creates software virtual displays so visual AI agents can work in desktop apps while physical monitors stay off.
- Stars
- 162
- Forks
- 12
- Open issues
- 8
- Last commit
- 3 days ago
What is it?
yomado is a native macOS app that creates software virtual displays for visual agents. Agents such as Codex read screenshots and operate desktop applications with the mouse and keyboard, which is useful for apps like Blender or Marvelous Designer that lack a suitable MCP interface. yomado supplies a display surface for those windows, so the graphical desktop stays available without a hardware dummy plug.
You create named virtual screens (1920 × 1080 or 2560 × 1440, optional HiDPI) and drag application icons between them and physical displays. The app can also move windows of a running app selected from Finder, and it returns windows to their original placement when moved back. A bundled skill lets a computer-use-capable agent set up and manage these screens for a task.
Virtual displays belong to the existing login session and share keyboard, pointer and focus with the user. The project relies on private macOS APIs, so compatibility may change with system updates, and monitor standby and unattended operation are not validated.
Key features
- 01Creates multiple named virtual displays at 1920 × 1080 or 2560 × 1440, with optional HiDPI.
- 02Drag app icons between virtual and physical screens, or move windows manually.
- 03Moves a running app's windows into a virtual screen from a .app bundle chosen in Finder.
- 04Restores windows to their original placement when returned to a physical display.
- 05Includes a skill that lets a computer-use agent set up and manage work screens.
- 06Keeps the system and desktop awake while virtual displays exist.
Use cases
- Letting Codex operate Blender or Marvelous Designer through screenshots and input.
- Giving a visual agent its own workspace while the user keeps working on physical monitors.
- Keeping a graphical desktop available while physical monitors are switched off manually.
Quick start
- 1
Clone the repository
- 2
Build the app from source
bash ./build.sh - 3
Open the built app at dist/Yomado.app
Things to know
- Requires macOS; no Windows or Linux support.
- Uses private macOS APIs, so compatibility can change with system updates.
- Experimental prototype; monitor-off operation and long unattended use are not validated.
- Virtual displays share keyboard, mouse and focus with your own session.
Topics
FAQ
- What is yomado?
- A native macOS app that creates software virtual displays so visual AI agents can work in desktop apps while physical monitors stay off. yomado is a native macOS app that creates software virtual displays for visual agents. Agents such as Codex read screenshots and operate desktop applications with the mouse and keyboard, which is useful for apps like Blender or Marvelous Designer that lack a suitable MCP interface. yomado supplies a display surface for those windows, so the graphical desktop stays available without a hardware dummy plug.
- Is yomado free to use? What license does it have?
- yomado is open source under the MIT license. Check the license terms before using it in commercial products.
- Is yomado still maintained?
- Actively maintained. The last commit to yomado was on Oct 17, 2026.
- What can I use yomado for?
- Letting Codex operate Blender or Marvelous Designer through screenshots and input. Giving a visual agent its own workspace while the user keeps working on physical monitors. Keeping a graphical desktop available while physical monitors are switched off manually.
- How do I install yomado?
- Clone the repository Build the app from source: bash ./build.sh Open the built app at dist/Yomado.app
- Where can I get yomado?
- The source code and installation instructions are on GitHub: https://github.com/chekusu/yomado
Summary generated by AI from the project's README. GitHub stats refresh daily.
Related projects
An open-source Computer Use service exposed as an MCP server, letting AI agents control macOS, Linux, and Windows desktops.
Lets AI agents like Claude Code or Codex read the screen, tap, type and swipe on a real iPhone over USB or Wi-Fi via MCP.
A free, open-source collection of native macOS menu bar tools with customizable panels, quick controls, automation, and plugins.
A browser built for AI agents and you to share: agents run tasks in isolated Spaces while your own tabs and logins stay untouched.
A DSH plugin that lets an AI model drive a running Blender session over a direct TCP channel, exposing 16 tools for viewport capture, scene edits and headless work.