gamedev-mcp-server
Curated game development knowledge for AI coding tools. 144 docs covering MonoGame, Godot, game design, and programming patterns.
README
GameDev MCP Server
Your AI forgets everything mid-project. Give it permanent game development knowledge.
GameDev MCP Server is a knowledge layer for AI coding assistants. It provides 140+ curated game development docs — design patterns, architecture guides, engine-specific implementation details — delivered through MCP so your AI assistant never loses context on how to build games.
Works with Claude Code, Claude Desktop, Cursor, Windsurf, Cline, and any MCP-compatible tool.
The Problem
Every game dev using AI hits the same wall: your assistant starts strong, then forgets your architecture mid-session. It suggests deprecated APIs. It doesn't know the difference between a state machine and a behavior tree. It writes Unity 5 code when you're on Unity 6.
GameDev MCP Server solves this by giving your AI a persistent, searchable knowledge base of curated game dev expertise — not raw docs, but structured implementation guidance that actually helps you build.
What's Inside
| Category | Examples | Docs |
|---|---|---|
| 🎮 Game Design | Genre systems, game feel, balancing, progression | 12 |
| 🏗️ Architecture | ECS, state machines, scene management, signals | 18 |
| 💻 Programming | Design patterns, data structures, algorithms | 15 |
| 🎯 Engine Guides | MonoGame (78 guides), Godot (11 docs), Unity (planned) | 89+ |
| 🔧 Core Concepts | Camera, physics, pathfinding, networking, combat, particles, UI | 19 |
| 📋 Project Mgmt | Scope control, sprint planning, art pipeline | 7 |
140+ docs. 4MB+ of curated knowledge. Zero external dependencies.
Quick Start
npx gamedev-mcp-server
That's it. No install required. Add it to your MCP config and your AI has instant game dev knowledge.
Claude Code
claude mcp add gamedev -- npx -y gamedev-mcp-server
Claude Desktop / Cursor / Windsurf / Cline
Add to your MCP config file:
{
"mcpServers": {
"gamedev": {
"command": "npx",
"args": ["-y", "gamedev-mcp-server"]
}
}
}
Config file locations:
- Claude Desktop:
claude_desktop_config.json - Cursor:
.cursor/mcp.json - Windsurf:
~/.windsurf/mcp.json - Cline: VS Code settings → Cline MCP Servers
Engine Modules
GameDev MCP Server uses a modular architecture. Core knowledge (design, patterns, algorithms) is always available. Engine-specific modules add implementation guides for your stack.
| Module | Status | Docs | Description |
|---|---|---|---|
core |
✅ Stable | 52 | Engine-agnostic game dev knowledge |
monogame-arch |
✅ Stable | 78 | MonoGame + Arch ECS — guides G1–G69, architecture, library reference |
godot-arch |
🚧 Active | 11 | Godot 4.4+ — architecture, GDScript/C#, scene composition, state machines, signals, input, physics, camera, tilemaps, animation |
unity-arch |
📋 Planned | — | Unity 6 — URP, ECS, modern patterns |
bevy-arch |
📋 Planned | — | Bevy ECS — Rust game dev |
Modules are auto-discovered. To filter which modules load:
{
"env": {
"GAMEDEV_MODULES": "monogame-arch,godot-arch"
}
}
Without GAMEDEV_MODULES, all available modules load automatically.
MCP Tools
| Tool | Description |
|---|---|
| Tool | Description |
| ------ | ------------- |
search_docs |
Full-text search across all docs with TF-IDF ranking, category/module/engine filters, cross-engine grouping |
get_doc |
Fetch a doc by ID with optional section extraction and maxLength for context efficiency |
list_docs |
Browse docs by category and module, with compact summary mode |
list_modules |
Discover available engine modules and their status |
compare_engines |
Compare how different engines approach the same topic (e.g., camera, physics, input) |
random_doc |
Discover docs serendipitously — great for exploration and learning |
genre_lookup |
Genre → required systems mapping (platformer, roguelike, tower defense, etc.) |
session |
Dev session co-pilot — structured workflows for planning, debugging, scoping |
license_info |
Show current tier and features |
Context-Efficient by Design
Unlike tool-heavy MCP servers that dump 50K+ tokens of schemas into your context window, GameDev MCP Server is built for precision:
- Section extraction —
get_doc("G64", section: "Knockback")returns just the knockback section, not the full 52KB doc maxLengthparam — Cap any response to fit your context budget- 9 focused tools — Minimal schema overhead, maximum utility. Compare to Godot MCP servers with 95+ tools burning half your context on schema alone
- stdio transport — No network exposure, no attack surface (MCP security is a real concern)
Free vs Pro
The server works fully out of the box with a generous free tier.
| Feature | Free | Pro |
|---|---|---|
| Core docs (design, patterns, algorithms) | ✅ | ✅ |
| Search | Core docs | All modules |
| Engine modules (MonoGame, Godot, etc.) | — | ✅ |
| Session co-pilot | — | ✅ |
| Genre lookup | Summary | Full details |
get_doc section extraction |
✅ | ✅ |
Pro unlocks engine-specific implementation guides — the stuff that turns "I know game design theory" into "here's exactly how to build it in Godot/MonoGame/Unity."
Get a Pro license → gamedev-mcp.lemonsqueezy.com
License Setup
Add your key to the MCP config:
{
"env": {
"GAMEDEV_MCP_LICENSE": "your-license-key"
}
}
Or create ~/.gamedev-mcp/license.json:
{ "key": "your-license-key" }
The server validates on startup, caches for 24h, and gracefully falls back to free tier if anything goes wrong. It never crashes.
What Makes This Different
There are 14,000+ MCP servers out there. Here's why this one matters for game dev:
- Knowledge, not integration. Godot-MCP, Unity-MCP, and Unreal-MCP give your AI buttons to press in the editor. This gives your AI understanding of how to architect and build games. They're complementary — use both.
- Cross-engine. One server, multiple engines. Learn a pattern once in core theory, then get the engine-specific implementation.
compare_engines("camera")shows you how Godot and MonoGame each handle it. No need to install separate MCPs per engine. - Curated, not scraped. Every doc is hand-written with AI code generation in mind — typed examples, anti-pattern warnings, decision trees, and "when to use" guidance. This isn't a docs mirror.
- Secure by design. stdio-only transport — no network exposure, no open ports, no attack surface. While 7,000+ MCP servers sit exposed on the internet, this runs entirely local.
- Grows with you. New docs and engines added continuously. Your AI gets smarter over time without you changing anything.
Genre Coverage
The genre_lookup tool maps any genre to its required systems with implementation priorities:
Platformer · Metroidvania · Roguelike · Tower Defense · Survival · RPG · Bullet Hell · Top-Down Shooter · Side-Scrolling · Fighting · Puzzle
Each genre profile includes: required systems, optional enhancements, suggested doc reading order, and a starter checklist.
Development
git clone https://github.com/sbenson2/gamedev-mcp-server.git
cd gamedev-mcp-server
npm install
npm run build
npm test # 187 tests, Node.js built-in test runner
npm run dev # Watch mode
Dev Mode
Skip license validation for local development:
{
"env": {
"GAMEDEV_MCP_DEV": "true"
}
}
Doc Structure
docs/
├── core/ # Engine-agnostic (always loaded)
│ ├── game-design/ # Genre profiles, game feel, balancing
│ ├── programming/ # Patterns, principles, data structures
│ ├── concepts/ # Camera, physics, pathfinding, networking, particles
│ ├── project-management/ # Scope, sprints, pipelines
│ ├── ai-workflow/ # AI code generation best practices
│ └── session/ # Co-pilot workflow prompts
├── monogame-arch/ # MonoGame + Arch ECS
│ ├── architecture/ # ECS overview, migration notes
│ ├── guides/ # G1–G69 implementation guides
│ └── reference/ # Library stack, project structure
└── godot-arch/ # Godot 4.4+
├── architecture/ # Node tree philosophy, GDScript vs C# decision guide
├── guides/ # Scene composition, state machines, signals, input, physics, camera, tilemaps, animation
└── reference/ # (coming soon)
MCP Resources
Docs are also available as MCP resources for clients that support them:
gamedev://docs/{module}/{id}— Any doc by module and IDgamedev://prompts/session— Session co-pilot promptgamedev://prompts/code-rules— AI code generation rules
Contributing
Found a bug? Have a doc suggestion? Open an issue.
License
MIT — see LICENSE.
Built for game devs who use AI. Stop fighting context loss. Start building.
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。
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。
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。
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。