DuckDB MCP Server

DuckDB MCP Server

A local MCP server enabling AI assistants to query and analyze data via DuckDB SQL engine, supporting local files, memory, S3, and MotherDuck.

Category
Visit Server

README

DuckDB MCP Server

npm version npm downloads npm license GitHub stars npm total downloads

已改名为 mcp-duckdb-local

📖 目录


📝 简介

这是一个本地运行的 Model Context Protocol (MCP) 服务器,允许 AI 助手通过 DuckDB 的强大分析 SQL 引擎访问数据。

🆕 新版本更新

v0.4.0 🚀: 项目重命名

为避免与 npm 上存在的包名冲突,本项目已改用无 scope 的包名 mcp-duckdb-local,可直接发布到 npmjs.com 公共仓库。

✨ 特点

功能 描述
SQL 查询 使用 DuckDB SQL 语法执行读/写查询
目录工具 浏览数据库、表和列
多连接支持 连接到本地文件、内存、S3 或 MotherDuck
读/写模式 可配置的读/写访问控制
安全模式 SaaS 模式限制本地文件系统访问
查询限制 最大行数和字符数限制
查询超时 防止长时间运行的查询
初始化 SQL 启动时执行自定义 SQL

🚀 快速开始

前置要求

  • Node.js 18+ 或更高版本
  • yarn >= 4

安装

# 使用 npm
npm install mcp-duckdb-local

# 或使用 pnpm
pnpm add mcp-duckdb-local

# 或使用 yarn
yarn add mcp-duckdb-local

# 使用 npx (无需安装)
npx mcp-duckdb-local --db-path :memory: --read-write

# 使用 yarn dlx (无需安装)
yarn dlx mcp-duckdb-local --db-path :memory: --read-write

在内存模式下运行 (开发模式)

{
  "mcpServers": {
    "DuckDB (内存模式)": {
      "command": "npx",
      "args": ["-y", "mcp-duckdb-local@latest", "--db-path", ":memory:", "--read-write"]
    }
  }
}

注意: Windows 用戶建議使用全局安裝(見下方),避免 npx 執行問題。

連接到本地 DuckDB 文件 (只讀模式)

{
  "mcpServers": {
    "DuckDB (只读)": {
      "command": "npx",
      "args": ["-y", "mcp-duckdb-local@latest", "--db-path", "/absolute/path/to/your.duckdb"]
    }
  }
}

連接到 MotherDuck (读写模式)

{
  "mcpServers": {
    "MotherDuck": {
      "command": "npx",
      "args": ["-y", "mcp-duckdb-local@latest", "--db-path", "md:", "--read-write"],
      "env": {
        "MOTHERDUCK_TOKEN": "<YOUR_TOKEN>"
      }
    }
  }
}

⚠️ Windows 用戶特別說明

在 Windows 上使用 npx 可能會遇到執行問題(Windows Script Host 錯誤)。建議使用以下任一方案:

方案 1:全局安裝(推薦)

npm install -g mcp-duckdb-local

然後在 MCP 配置中使用:

{
  "mcpServers": {
    "DuckDB": {
      "command": "mcp-duckdb-local",
      "args": ["--db-path", "D:/path/to/db.duckdb"]
    }
  }
}

方案 2:使用 npx 時添加 -y 標誌

{
  "mcpServers": {
    "DuckDB": {
      "command": "npx",
      "args": ["-y", "mcp-duckdb-local@latest", "--db-path", "D:/path/to/db.duckdb"]
    }
  }
}

方案 3:使用 yarn dlx

{
  "mcpServers": {
    "DuckDB": {
      "command": "yarn",
      "args": ["dlx", "mcp-duckdb-local", "--db-path", "D:/path/to/db.duckdb"]
    }
  }
}

📦 安装

全局安装

npm install -g mcp-duckdb-local
# 或
pnpm add -g mcp-duckdb-local
# 或
yarn global add mcp-duckdb-local

本地安装

npm install mcp-duckdb-mcp --save
# 或
pnpm add mcp-duckdb-local
# 或
yarn add mcp-duckdb-local

📖 使用方法

CLI 命令

# 显示帮助
npx -y mcp-duckdb-local@latest --help

# 連接到本地文件 (只读)
npx -y mcp-duckdb-local@latest --db-path /path/to/database.duckdb

# HTTP 服务器模式
npx -y mcp-duckdb-local@latest --transport http --port 8080

# 全局安裝後直接使用
mcp-duckdb-local --db-path :memory: --read-write

命令行参数

参数 默认值 说明
--db-path :memory: 数据库路径 (本地文件, :memory:, md:, s3://)
--motherduck-token env variable MotherDuck 访问令牌
--read-write false 启用写访问
--motherduck-saas-mode false MotherDuck SaaS 模式
--allow-switch-databases false 启用数据库切换工具
--max-rows 1024 最大返回行数
--max-chars 50000 最大返回字符数
--query-timeout -1 查询超时 (秒), -1 禁用
--init-sql none 启动时执行的 SQL
--transport stdio 传输类型: stdiohttp
--host 127.0.0.1 HTTP 主机
--port 8000 HTTP 端口
--home-dir system HOME DuckDB 主目录
--ephemeral-connections true 对本地文件使用临时连接

环境变量

变量 说明
MOTHERDUCK_TOKEN MotherDuck 访问令牌
MCP_DB_PATH 数据库路径
MCP_TRANSPORT 传输类型
MCP_PORT HTTP 端口
MCP_HOST HTTP 主机
MCP_READ_WRITE 启用写访问
MCP_SAAS_MODE SaaS 模式
MCP_ALLOW_SWITCH_DATABASES 启用数据库切换
MCP_MAX_ROWS 最大行数
MCP_MAX_CHARS 最大字符数
MCP_QUERY_TIMEOUT 查询超时
HOME DuckDB 主目录

🔧 工具

execute_query

执行 SQL 查询。

参数:

  • sql (string, 必需): 要执行的 SQL 查询

示例:

{
  "name": "execute_query",
  "arguments": {
    "sql": "SELECT * FROM my_table LIMIT 10"
  }
}

返回:

{
  "success": true,
  "columns": ["id", "name", "created_at"],
  "columnTypes": ["BIGINT", "VARCHAR", "TIMESTAMP"],
  "rows": [
    [1, "Alice", "2024-01-01"],
    [2, "Bob", "2024-01-02"]
  ],
  "rowCount": 2
}

list_databases

列出所有可用数据库。

参数: 无

示例:

{
  "name": "list_databases",
  "arguments": {}
}

返回:

{
  "success": true,
  "databases": [
    { "name": "memory", "type": "local" },
    { "name": "main", "type": "local" }
  ],
  "databaseCount": 2
}

list_tables

列出数据库中的所有表和视图。

参数:

  • database (string, 可选): 数据库名称 (默认: 当前数据库)
  • schema (string, 可选): 模式名称 (默认: 所有模式)

示例:

{
  "name": "list_tables",
  "arguments": {
    "database": "main",
    "schema": "main"
  }
}

返回:

{
  "success": true,
  "database": "main",
  "schema": "main",
  "tables": [
    {
      "schema": "main",
      "name": "users",
      "type": "table",
      "comment": "User accounts"
    },
    {
      "schema": "main",
      "name": "active_users",
      "type": "view",
      "comment": null
    }
  ],
  "tableCount": 1,
  "viewCount": 1
}

list_columns

列出表或视图的所有列。

参数:

  • table (string, 必需): 表或视图名称
  • database (string, 可选): 数据库名称
  • schema (string, 可选): 模式名称

示例:

{
  "name": "list_columns",
  "arguments": {
    "table": "users",
    "database": "main"
  }
}

返回:

{
  "success": true,
  "database": "main",
  "schema": "main",
  "table": "users",
  "objectType": "table",
  "columns": [
    {
      "name": "id",
      "type": "BIGINT",
      "nullable": false,
      "comment": null
    },
    {
      "name": "name",
      "type": "VARCHAR",
      "nullable": true,
      "comment": "User name"
    }
  ],
  "columnCount": 2
}

switch_database_connection

切换到不同的数据库 (可选启用)。

参数:

  • path (string, 必需): 数据库路径
  • create_if_not_exists (boolean, 可选): 如果不存在则创建 (仅读写模式)

示例:

{
  "name": "switch_database_connection",
  "arguments": {
    "path": "/path/to/new/database.duckdb"
  }
}

返回:

{
  "success": true,
  "message": "Switched to database: /path/to/new/database.duckdb",
  "previousDatabase": ":memory:",
  "currentDatabase": "/path/to/new/database.duckdb",
  "readOnly": false
}

⚙️ 配置

Claude Desktop

Settings → Developer → Edit Config

{
  "mcpServers": {
    "DuckDB (内存)": {
      "command": "npx",
      "args": ["-y", "mcp-duckdb-local@latest", "--db-path", ":memory:", "--read-write"]
    }
  }
}

Windows 用戶: 建議全局安裝後使用 mcp-duckdb-local 命令替代 npx

Claude Code

claude mcp add --scope user duckdb \
  --transport stdio \
  -- npx -y mcp-duckdb-local@latest --db-path :memory: --read-write

VS Code

Ctrl+Shift+P → Preferences: Open User Settings (JSON)

{
  "mcpServers": {
    "DuckDB": {
      "command": "npx",
      "args": ["-y", "mcp-duckdb-local@latest", "--db-path", "/path/to/db.duckdb"]
    }
  }
}

Cursor

Settings → MCP → Add new global MCP server


🧑 开发文档

详细的开发指南、构建命令、测试和代码规范已移至 DEVELOPMENT.md

快速参考:

# 安装依赖
yarn install

# 开发模式
yarn dev

# 构建
yarn build

# 运行测试
yarn test

# 打包分发
yarn build:bundle

完整开发文档请查看 DEVELOPMENT.md


🤝 贡献

我们欢迎贡献!请遵循以下步骤:

  1. Fork 仓库
  2. 创建功能分支 (git checkout -b feature/amazing-feature)
  3. 提交更改 (git commit -m 'feat: add amazing feature')
  4. 推送到分支 (git push origin feature/amazing-feature)
  5. 提交 Pull Request

贡献者


📄 许可证

本项目采用 Apache-2.0 许可证


🙏 致谢


📞 支持


本文档由 DuckDB MCP Server 生成 🦆

Recommended Servers

playwright-mcp

playwright-mcp

A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.

Official
Featured
TypeScript
Magic Component Platform (MCP)

Magic Component Platform (MCP)

An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.

Official
Featured
Local
TypeScript
Audiense Insights MCP Server

Audiense Insights MCP Server

Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.

Official
Featured
Local
TypeScript
VeyraX MCP

VeyraX MCP

Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.

Official
Featured
Local
graphlit-mcp-server

graphlit-mcp-server

The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.

Official
Featured
TypeScript
Kagi MCP Server

Kagi MCP Server

An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.

Official
Featured
Python
E2B

E2B

Using MCP to run code via e2b.

Official
Featured
Neon Database

Neon Database

MCP server for interacting with Neon Management API and databases

Official
Featured
Exa Search

Exa Search

A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.

Official
Featured
Qdrant Server

Qdrant Server

This repository is an example of how to create a MCP server for Qdrant, a vector search engine.

Official
Featured