Godot MCP Documentation Server
Provides AI assistants with direct access to the complete Godot Engine documentation, including classes, tutorials, and features. It enables developers to retrieve and analyze official documentation through natural language interfaces using the Model Context Protocol.
README
Godot MCP Documentation Server
中文文档 | English
A Model Context Protocol (MCP) server that provides AI assistants with access to the complete Godot Engine documentation, helping developers with Godot development by serving documentation directly to LLMs.
Purpose
This server bridges the gap between AI assistants and Godot documentation, allowing developers to get instant, accurate answers about Godot classes, tutorials, and features without leaving their AI chat interface.
Project Origin
This project is based on godot-mcp-docs by Nihilantropy, modified for local development on Windows without Docker.
Installation
-
Install uv (if not already installed):
# On Windows with PowerShell: Invoke-WebRequest -Uri https://astral.sh/uv/install.ps1 -UseBasicParsing | Invoke-Expression -
Clone the repository:
git clone https://github.com/Nihilantropy/godot-mcp-docs.git cd godot-mcp-docs -
Install Python dependencies:
uv syncThis command reads from
pyproject.tomland creates a virtual environment.Note:
uv syncusespyproject.tomlas the source of truth, notrequirements.txt. This is the modern Python packaging standard (PEP 517/518/621). Theuv.lockfile ensures consistent dependency versions across environments.
Setting Up Documentation
After installing dependencies, you need to download and process the Godot documentation:
-
Generate documentation:
uv run python .\docs_converter\godot_docs_converter.py -
Move docs folder to root: After conversion, move the generated
docsfolder to the project root directory. -
Generate documentation tree:
cd docs tree /f > docs_tree.txt cd ..Note: If you encounter Chinese encoding issues, use:
tree /f | Out-File -Encoding utf8 docs_tree.txt
Running the Server
Start the MCP server locally:
uv run python main.py
Configuring MCP Client
Claude Desktop Example
Add this to your Claude Desktop configuration file:
{
"mcpServers": {
"godot-mcp-docs": {
"command": "uv",
"args": [
"run",
"python",
"main.py"
]
}
}
}
Available Tools
get_documentation_tree()- Get a tree-style overview of the entire documentation structureget_documentation_file(file_path: str)- Retrieve the content of specific documentation files
Sample Usage
Explore documentation structure:
What documentation is available for Godot?
Get specific class documentation:
Show me the documentation for CharacterBody2D
Learn about tutorials:
What tutorials are available for 2D game development?
Get specific tutorial content:
Show me the first 2D game tutorial
Compare classes:
What's the difference between Node2D and CharacterBody2D?
Debugging
Option 1: FastMCP Dev UI (Recommended)
Start the server in dev mode:
uv run fastmcp dev main.py
Then open your browser to http://127.0.0.1:8000 to see all tools and test them interactively.
Option 2: MCP Inspector (Official Debugger)
Use Anthropic's official MCP Inspector:
npx @modelcontextprotocol/inspector uv run python main.py
This opens http://localhost:5173 in your browser, showing all tools, request/response logs, and allowing manual tool testing. No need to run uv run python main.py separately.
Option 3: HTTP API Debugging
Send JSON-RPC requests to the server:
Invoke-WebRequest -Uri http://localhost:8000/message `
-Method POST `
-ContentType "application/json" `
-Body '{
"jsonrpc": "2.0",
"method": "tools/call",
"params": {
"name": "get_documentation_file",
"arguments": {
"file_path": "classes/class_characterbody2d.md"
}
},
"id": 1
}'
Or using curl:
curl -X POST http://localhost:8000/message `
-H "Content-Type: application/json" `
-d '{
"jsonrpc": "2.0",
"method": "tools/call",
"params": { "name": "search_godot_docs", "arguments": { "query": "DisplayServer" } },
"id": 1
}'
Understanding Path Resolution
The DOCS_DIR = Path("docs").resolve() in the code is relative to the current working directory (CWD) when you run Python, not the script location.
Example
Assume your directory structure:
D:\Codes\16_MCP\ <-- 根目录
├── main.py <-- 启动脚本
├── docs\ <-- 文档目录
└── srcs\
└── util\
└── docs_utils.py <-- 工具代码 (里面写了 Path("docs"))
Case A: Running from root directory
cd D:\Codes\16_MCP
python main.py
- Current Working Directory:
D:\Codes\16_MCP Path("docs")resolves to:D:\Codes\16_MCP\docs(Success)
Case B: Running from subdirectory (common error)
cd D:\Codes\16_MCP\srcs
python ../main.py
- Current Working Directory:
D:\Codes\16_MCP\srcs Path("docs")resolves to:D:\Codes\16_MCP\srcs\docs(Fails - no docs folder)
Best Practice
To ensure paths work regardless of where you run from, use __file__-based absolute positioning:
from pathlib import Path
# Get current script directory
CURRENT_SCRIPT_DIR = Path(__file__).resolve().parent
# Navigate to project root
PROJECT_ROOT = CURRENT_SCRIPT_DIR.parent.parent
# Locate docs directory
DOCS_DIR = (PROJECT_ROOT / "docs").resolve()
PowerShell Output Redirection Tips
Save output to file:
-
Overwrite (clears existing content):
ls > list.txt -
Append (preserves existing content):
ls >> list.txt -
With UTF8 encoding (recommended for Chinese):
ls | Out-File -FilePath output.txt -Encoding utf8 -
See on screen and save:
ls | Tee-Object -FilePath log.txt
Documentation Structure
The server provides access to the complete official Godot documentation with this structure:
docs/
├── _styleguides
├── _tools
│ └── redirects
├── about
├── classes
├── community
│ └── asset_library
├── contributing
│ ├── development
│ │ ├── compiling
│ │ ├── configuring_an_ide
│ │ ├── core_and_modules
│ │ ├── debugging
│ │ │ └── vulkan
│ │ ├── editor
│ │ └── file_formats
│ ├── documentation
│ └── workflow
├── getting_started
│ ├── first_2d_game
│ ├── first_3d_game
│ ├── introduction
│ └── step_by_step
├── img
└── tutorials
├── 2d
├── 3d
│ ├── global_illumination
│ ├── particles
│ └── procedural_geometry
├── animation
├── assets_pipeline
│ ├── escn_exporter
│ └── importing_3d_scenes
├── audio
├── best_practices
├── editor
├── export
├── i18n
├── inputs
├── io
├── math
├── migrating
├── navigation
├── networking
├── performance
│ └── vertex_animation
├── physics
│ └── interpolation
├── platform
│ ├── android
│ ├── ios
│ └── web
├── plugins
│ └── editor
├── rendering
├── scripting
│ ├── c_sharp
│ │ └── diagnostics
│ ├── cpp
│ ├── debug
│ ├── gdextension
│ └── gdscript
├── shaders
│ ├── shader_reference
│ └── your_first_shader
├── ui
└── xr
Recommended System Prompt
For optimal results when working with Godot, use this system prompt:
"When working with Godot game development questions, always search for the latest available documentation using the godot-mcp-docs tools. Start with
get_documentation_tree()to understand the documentation structure, then useget_documentation_file()to retrieve specific information about classes, tutorials, or features. Prioritize official Godot documentation over general knowledge when providing Godot-related assistance."
Updating Documentation
To update to a newer version of Godot documentation:
uv run python .\docs_converter\godot_docs_converter.py
cd docs
tree /f > docs_tree.txt
License
This project is licensed under the MIT License - see the LICENSE file for details.
The Godot documentation content follows the original Godot documentation licensing:
- Documentation content (excluding
classes/folder): CC BY 3.0 - Class reference files (
classes/folder): MIT License - Attribution: "Juan Linietsky, Ariel Manzur and the Godot community"
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。