Paper page - OpenGame: Open Agentic Coding for Games

Source: https://huggingface.co/papers/2604.18394

Abstract

OpenGame is an open-source agentic framework for end-to-end web game creation that uses specialized code models and evaluation benchmarks to overcome challenges in interactive application development.

Game development sits at the intersection of creative design and intricate software engineering, demanding the joint orchestration ofgame engines,real-time loops, and tightly coupled state across many files. WhileLarge Language Models(LLMs) andcode agentsnow solve isolated programming tasks with ease, they consistently stumble when asked to produce a fully playable game from a high-level design, collapsing undercross-file inconsistencies, brokenscene wiring, andlogical incoherence. We bridge this gap with OpenGame, the first open-source agentic framework explicitly designed for end-to-end web game creation. At its core liesGame Skill, a reusable, evolving capability composed of aTemplate Skillthat grows a library of project skeletons from experience and aDebug Skillthat maintains a living protocol of verified fixes - together enabling the agent to scaffold stable architectures and systematically repair integration errors rather than patch isolated syntax bugs. Powering this framework isGameCoder-27B, a code LLM specialized for game engine mastery through a three-stage pipeline ofcontinual pre-training,supervised fine-tuning, andexecution-grounded reinforcement learning. Since verifying interactive playability is fundamentally harder than checking static code, we further introduceOpenGame-Bench, an evaluation pipeline that scores agentic game generation alongBuild Health,Visual Usability, andIntent Alignmentviaheadless browser executionandVLM judging. Across 150 diverse game prompts, OpenGame establishes a new state-of-the-art. We hope OpenGame pushescode agentsbeyond discrete software engineering problems and toward building complex, interactive real-world applications. Our framework will be fully open-sourced.

View arXiv pageView PDFProject pageGitHub2.03kAdd to collection

Get this paper in your agent:

hf papers read 2604\.18394

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper0

No model linking this paper

Cite arxiv.org/abs/2604.18394 in a model README.md to link it from this page.

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2604.18394 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2604.18394 in a Space README.md to link it from this page.

Collections including this paper16

Browse 16 collections that include this paper