As far as I understand the current implementation for RealityKit/ARKit, you cannot auto-anchor Window Groups / windows / volumes to detected surfaces. You can only anchor entities inside of an immersive space. (you can however use mixed or full immersion, but you cannot use the shared space) This means, any auto-anchoring by the app will eliminate the possibility to multi-task for the user.

I stumbled upon the same limitations. I decided I'll just make two versions of my app, leave the choice up to the user: have them choose a volume, and they'll need to put it somewhere but can multi-task have them choose a full space, and my app will auto anchor to their table, but they cannot multi-task anymore I don't think there's much more we can do with these current limitations. Maybe next year things will be improved and limitations might get lifted. In the simulator, it seems that when I drag the volume around to try to place it on a surface, geometry inside of a RealityView can clip through “real” objects. Is this the expected behavior on a real device too? as far as my understanding goes, I think yes, this is the expected behaviour If so, could using ARKit in a Full Space to position the volume, then switching back to a Shared Space, be an option? as far as my understanding goes, I think no, because I don't think we can anchor volumetric windows. that's the main limitation. So you'd have to work with just plain entities that you anchor, but then switching back to a volume will make it so those entities disappear. Also, if the app is closed, and reopened, will the volume maintain its position relative to the user’s real-world environment? I think the volume will stay where the user put it yes, but this is at apple's regression on how they choose to implement the behaviour of volumes in the shared space. it might change at any time and developers won't be able to have a say in it. Think of it as, we don't know where on a macbook screen and what size a window is exactly located at. Same for a volumetric window.