mcp-codebase-index

概述

mcp-codebase-index 是一个结构化代码库索引器，将源文件解析为元数据并通过 MCP 提供 18 个查询工具。它支持 Python、TypeScript/JS、Go、Rust 和 C# 文件，在 git 仓库中自动进行增量重新索引。服务器使用持久磁盘缓存来消除冷启动延迟，并能高效处理大型代码库（已在拥有 110 万行代码的 CPython 上测试）。

试试问 AI

装完之后，这里有 6 个你可以让 AI 做的事：

• get_project_summary

  Get file count, packages, top classes/functions

  试试:Show me the project overview试试:What are the main components of this codebase?
• list_files

  List indexed files with optional glob filter

  试试:List all Python files试试:Show me all files in the src directory
• get_structure_summary

  Get structure of a file or the whole project

  试试:What's the structure of main.py?试试:Show me the overall project structure
• get_functions

  List functions with name, lines, params

  试试:List all functions试试:What functions are in utils.py?
• get_classes

  List classes with name, lines, methods, bases

  试试:Show me all classes试试:What classes exist in models.py?
• get_imports

  List imports with module, names, line

  试试:What modules are imported in main.py?试试:List all imports in the project
• get_function_source

  Get full source of a function/method

  试试:Show me the source of process_data试试:Get the implementation of calculate
• get_class_source

  Get full source of a class

  试试:Show me the User class implementation试试:Get the source code for Database
• find_symbol

  Find where a symbol is defined (file, line, type)

  试试:Where is authenticate defined?试试:Find the location of logger
• get_dependencies

  Get what a symbol calls/uses

  试试:What does process_data depend on?试试:Show dependencies of main
• get_dependents

  Get what calls/uses a symbol

  试试:What uses authenticate?试试:Show all callers of logger
• get_change_impact

  Get direct + transitive dependents of a symbol

  试试:What will break if I change User?试试:Show impact of modifying logger

可对比工具

安装

pip install "mcp-codebase-index[mcp]"

{
  "mcpServers": {
    "codebase-index": {
      "command": "mcp-codebase-index",
      "env": {
        "PROJECT_ROOT": "/path/to/project"
      }
    }
  }
}

FAQ

它如何高效处理大型代码库？

它使用持久磁盘缓存和基于 git diff 的增量更新，只重新解析已更改的文件（约 1-2 毫秒）。在 CPython（110 万行）上，初始索引构建需要 56 秒，但后续启动时从缓存加载不到 1 秒。

支持哪些编程语言？

Python（AST 解析）、TypeScript/JS、Go、Rust、C#（基于正则表达式）以及 Markdown/Text 文件。其他文件仅提供行数统计。

Hacker News 讨论

开发者社区最近的相关讨论。

• Show HN: MCP Codebase Index – 87% fewer tokens when AI navigates your codebase▲ 1
  帖子 by localforthewin · 2026-02-17

mcp-codebase-index 对比