Want to build your own game?
Vibe is all you need.
An agent team that builds your game from natural language.
Pattern Recognition Laboratory, Nanjing University
We present VibeGame, an agent system for Prompt-to-Game Development that builds playable 2D games from natural language and iteratively refines them through follow-up requests. This task simultaneously demands gameplay design, software engineering, and multimodal asset creation. Unlike conventional software, game projects require code, assets, animations, and configurations to maintain mutual consistency across heterogeneous artifact types, and correctness can only be verified by actually running the game. These challenges are compounded by the systematic absence of game development practice knowledge in public training corpora. VibeGame addresses these issues through three coupled designs. The system is built around an Adversarial Agent Team centered on an orchestrator that coordinates multiple specialized agents, separating generation from review and challenging every output through static review and end-to-end runtime verification from multiple independent perspectives. This team operates on an AI-native game engine that enables agents to read and write all project state without GUI interaction and to verify runtime behavior synchronously. As projects are successfully delivered, a training-free self-evolution mechanism distills reusable project skeletons, functional modules, and collaboration contracts, continuously enriching the framework's built-in knowledge so that subsequent projects start from a higher baseline. We demonstrate the system's end-to-end capabilities through multiple games of different types, covering creation of complete playable projects from text-only or multimodal input, as well as editing scenarios including IP transfer, genre transfer, and rule restructuring on existing projects.
An orchestrator aligns intent, then drives parallel art, code, and design workflows. Each code sub-task passes through architect, programmer, auditor, and player, and an independent reviewer accepts the whole project on functionality, visual quality, and playability.
Scenes are plain JSON, assets register through one entry point, and logic stays separate from structure. Agents build the game by editing text files and play it at their own pace, stepping the runtime frame by frame.
Each delivered project is distilled into a reproducible genre skeleton, self-checking reusable modules, and collaboration contracts: accumulated experience that future projects start from.
