GameCodex
MCP server that provides game design lenses and knowledge base docs to help AI assistants and programmers make better design decisions during game development.
README
GameCodex
A game designer AI lens for programmers.
Your AI assistant writes competent game code and gives you zero design judgment.
It will implement a jump, a shop, or a crafting system without ever asking
whether it should exist, what it should feel like, or which proven design
thinking applies. GameCodex fixes exactly that: one MCP tool — lens —
that answers as the game designer looking over your shoulder.
you (or your AI): lens { situation: "my jump feels floaty" }
→ Game Feel (Steve Swink): measure input-to-photon latency;
derive gravity from jump height + time-to-apex; split
rising/falling gravity; the questions, red flags in code
terms, and concrete fixes.
Every lens distills an industry-proven philosophy with real provenance — Mark Cerny's Method, MDA, Sid Meier's interesting decisions, Swink's game feel, juice, flow & difficulty (incl. Celeste's assist thinking), George Fan's onboarding rules, Nintendo's kishōtenketsu, core loops, Derek Yu on finishing, Valve-style playtesting, Self-Determination Theory, Sirlin/Schreiber balance, Koster's theory of fun, and BotW-style emergence.
Install
claude mcp add gamecodex -- npx -y gamecodex
Or in any MCP config (claude_desktop_config.json, .cursor/mcp.json, …):
{
"mcpServers": {
"gamecodex": { "command": "npx", "args": ["-y", "gamecodex"] }
}
}
One tool — the whole game
| Call | Returns |
|---|---|
lens { situation: "players quit at the first boss" } |
The matching lenses (here: flow & difficulty, playtesting) — designer questions, red flags, prescriptions — plus matching docs from the knowledge base |
lens { lens: "scope" } |
One named lens in full |
lens { doc: "G106" } |
A knowledge-base doc (big ones return a table of contents; add section: "…" for one part) |
lens { } |
The catalog of all 15 lenses |
That's the entire tool surface. No schema bloat, no action routing.
The knowledge base works within the lens
The 957-doc engine library (29 engines — deepest on MonoGame, Godot, Unity,
Unreal) is reachable through the same tool: every situation reply ends with
the matching docs, and doc/section reads them — section-aware, so 100KB
guides never flood your context. The docs also ship as passive MCP
resources (gamedev://docs/...). Scope which modules load with
GAMEDEV_MODULES (e.g. core,godot-arch).
Monorepo
GameCodex/
├── packages/
│ ├── server/ <- the MCP server (the product) — see its README + SPEC.md
│ └── site/ <- marketing site (stale; predates the lens direction)
└── package.json <- npm workspaces root
npm install && npm run build && npm test
Links
- npm
- GitHub (canonical)
- Issues
- GitLab mirror
License
MIT
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。