Rooms

What Rooms does well is remove every step between an idea and a running 3D scene. It runs in the browser. There is no installer, no compile step, and no paid tier in the published pricing details, which lists a single Free plan at $0/mo. Publishing is a link: the site states plainly that you share your room link for playback on desktop and iOS, and the homepage nudges visitors to open rooms.xyz on desktop to build. The scripting surface is the real differentiator. The documentation at rooms.xyz/docs is organized the way a learner needs it — Introduction, Things and Methods, Special Functions, Handling Input, Physics Types, Shortcuts — and the method list spans far more than toy territory. You get sound control (playSound, playSoundWithPitch, setSoundVolume, setMusic, stopSound), physics (explode applies an explosion force centered on a given position, with a particles argument), particle effects (particles, stopParticles), math and strings, photo capture (takePhoto, instantPhoto, sharePhoto), comments (fetchComments, promptForComment), and high scores (loadHighScores, submitHighScore, deleteHighScore). Camera work is first-class too: setCamera changes the room camera to a given camera object, enterCameraMode opens a camera from the perspective of the current thing, and exitCameraMode closes it. isPortraitDevice lets you branch on phone-shaped screens. Where it gets genuinely interesting is persistence. setSessionValue stores a value that persists from room to room, with clearSessionValue, clearSession, and getSessionValue to manage it — that is what makes multi-room experiences and progression possible rather than a single disconnected scene. Separate from that, saveUserMemory and loadUserMemory store values per user, so a room can remember a returning visitor. High scores are their own system. Many people never use any of this. AI is presented as a primitive, not a bolt-on. aiCharacter is documented as the all-in-one AI conversation function, with aiChatCreate to open a conversation and aiChat to continue it with user input. Combined with the say function for speech bubbles, an AI character in a room reads as a character, not a text box. Multiplayer messaging exists but is deliberately minimal: send targets a specific thing, broadcast hits every thing in the room. The homepage shows follow counts and follower lists, so there is a social layer around rooms, but nothing in the scrape suggests real-time co-editing. Limitations are structural rather than bugs. Browser playback is stated for desktop and iOS; there is nothing in the evidence about native authoring clients on any platform. There is no sign of version control or a multi-user editing mode. There is no mention of dedicated support or SLAs. And because everything is browser-hosted, you do not control the runtime the way you would with Unity or Godot — a tradeoff no professional engine user would accept for a shipping title, and a perfectly reasonable one for a classroom. Where it