MaxCompute MCP Server

本项目是基于阿里云的MCP的修改版，对业务进行了加固以及迭代
阿里云项目地址: https://github.com/datafe/maxcompute-mcp-server
主要修改：
1. 增加了sqlglot的依赖，实现对sql的专业化解析和处理，而不是简单的字符串查找
2. 移除了describe_table这个tool。实际上AI是完全自己能知道使用desc来获得表描述，
   这个方法有点多此一举。
3. 相应的readme文案调整   

Overview

A Model Context Protocol (MCP) server for MaxCompute.

Components

Tools

The server offers 3 core tools:

Query Tools

• get_latest_partition

  • Get the latest partition name for a specific table
  • Input:
    • table_name (string): Name of the table
  • Returns: The latest partition name

• read_query

  • Execute a read-only query on the MaxCompute project, such as SELECT or DESC query is allowed.
  • Input:
    • query (string): The read only SQL query
    • result_format (string,optional): Output format. The selectable formats are 'csv' and 'json', with 'csv' being the default.
    • save_file_path (string,optional): local file that save the query results as json list
  • Returns: Query results as array of objects

Schema Tools

• list_tables
  • List all tables in the MaxCompute project
  • No input required
  • Returns: Array of table objects with name and comment

Build from source

• Code: GitLab

cd /path/to/tapdb-maxcompute-mcp
uv pip install .
uv build

Usage with Cline

• maxcompute endpoints can view in endpoint

• maxcompute projects can view in projects

# Add the server to your cline_mcp_settings.json
"mcpServers": {
  "tapdb-maxcompute-mcp": {
    "command": "uv",
    "args": [
      "run",
      "--with",
      "tapdb-maxcompute-mcp",
      "tapdb-maxcompute-mcp"
    ],
    "env": {
      "ALIBABA_CLOUD_ACCESS_KEY_ID": "",
      "ALIBABA_CLOUD_ACCESS_KEY_SECRET": "",
      "ALIBABA_CLOUD_MAXCOMPUTE_PROJECT": "",
      "ALIBABA_CLOUD_MAXCOMPUTE_ENDPOINT": "https://service.cn-shanghai.maxcompute.aliyun.com/api"
    }
  }
}