source-rag-mcp

source-rag-mcp

MCP server for searching, browsing, and analyzing decompiled Minecraft source code locally. Supports symbol lookup, text search, reference lookup, and lightweight RAG.

Category
访问服务器

README

source-rag-mcp

Minecraft decompiled source MCP server for local source search, symbol lookup, reference lookup, and lightweight RAG.

Copyright Boundary

This project does not ship Minecraft source code, bytecode, assets, jars, or decompiled output.

The package contains only the MCP server code. When you use add_minecraft_version or decompile_classes, Minecraft files are downloaded or generated only on your local machine under SOURCE_RAG_DATA or ./.source-rag.

Do not commit, publish, or redistribute:

  • .source-rag/
  • sources/
  • Minecraft .jar files
  • Minecraft .class files
  • decompiled Minecraft .java output

This project is licensed under Apache-2.0. Minecraft is owned by Mojang/Microsoft and is not included in this project.

Setup

pnpm install
pnpm build

Run

Run the server from the project root:

pnpm build
node ./dist/index.js

The server writes index data to ./.source-rag by default. You can override this location with the SOURCE_RAG_DATA environment variable.

First Index

For a normal Minecraft version, use add_minecraft_version. The MCP server will download the Minecraft jar from Mojang metadata, decompile it, and index it.

{
  "version": "latest_release",
  "side": "client"
}

Exact version IDs work too:

{
  "version": "26.2",
  "side": "client"
}

If you already have decompiled .java files, use index_sources with a version name and source directory:

{
  "version": "26.1",
  "sourceDir": "./sources/26.1"
}

If you have a jar, .class file, or class directory, use decompile_classes first:

{
  "input": "./sources/26.2",
  "outputDir": "./.source-rag/sources/26.2",
  "version": "26.2",
  "indexAfter": true
}

decompile_classes downloads Vineflower into ./.source-rag/tools on first use.

Tools

  • list_versions: list indexed source versions
  • add_minecraft_version: download a Minecraft jar from Mojang metadata, decompile it, and index it
  • index_sources: index a local decompiled Java source tree
  • decompile_classes: decompile class or jar input with Vineflower and optionally index it
  • search_symbol: search classes, methods, and fields
  • search_text: search raw source lines
  • rag_search: search source chunks with lightweight lexical scoring
  • get_source: read a source file by path or class name
  • get_method_source: read a method body from a class, including inner class owners and overloaded methods
  • find_references: find exact word references

get_method_source accepts these optional overload filters:

{
  "version": "26.2",
  "owner": "net.minecraft.world.item.ItemStack",
  "method": "ItemStack",
  "parameterTypes": ["Holder<Item>", "int"]
}

Inner classes can be addressed with either . or $:

{
  "version": "26.2",
  "owner": "com.mojang.blaze3d.vertex.TlsfAllocator.Block",
  "method": "isFree"
}

Codex MCP Config

Add this to your Codex config.toml. Prefer absolute paths because Codex may start the MCP server from a different working directory.

[mcp_servers.minecraft-source]
command = "node"
args = [
  "C:\\dev\\minecraft\\source-rag-mcp\\dist\\index.js"
]

[mcp_servers.minecraft-source.env]
SOURCE_RAG_DATA = "C:\\dev\\minecraft\\source-rag-mcp\\.source-rag"

If node is not on PATH, use an absolute Node executable path for command only:

[mcp_servers.minecraft-source]
command = "<path-to-node>"
args = [
  "C:\\dev\\minecraft\\source-rag-mcp\\dist\\index.js"
]

[mcp_servers.minecraft-source.env]
SOURCE_RAG_DATA = "C:\\dev\\minecraft\\source-rag-mcp\\.source-rag"

Example:

[mcp_servers.minecraft-source]
command = "C:/path/to/node.exe"
args = [
  "C:\\dev\\minecraft\\source-rag-mcp\\dist\\index.js"
]

[mcp_servers.minecraft-source.env]
SOURCE_RAG_DATA = "C:\\dev\\minecraft\\source-rag-mcp\\.source-rag"

Indexed Sources

Indexed source trees are stored under:

<SOURCE_RAG_DATA>/sources/<version>

For example:

C:\dev\minecraft\source-rag-mcp\.source-rag\sources\26.2

推荐服务器

Baidu Map

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。

官方
精选
JavaScript
Playwright MCP Server

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。

官方
精选
TypeScript
Magic Component Platform (MCP)

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。

官方
精选
本地
TypeScript
Audiense Insights MCP Server

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。

官方
精选
本地
TypeScript
VeyraX

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。

官方
精选
本地
graphlit-mcp-server

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。

官方
精选
TypeScript
Kagi MCP Server

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。

官方
精选
Python
e2b-mcp-server

e2b-mcp-server

使用 MCP 通过 e2b 运行代码。

官方
精选
Neon MCP Server

Neon MCP Server

用于与 Neon 管理 API 和数据库交互的 MCP 服务器

官方
精选
Exa MCP Server

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。

官方
精选