DuckDB MCP Server
A beginner-friendly MCP server that exposes a run_sql tool for querying a DuckDB employee database, returning results as Markdown tables. It enables AI assistants to run SQL queries through the Model Context Protocol.
README
DuckDB MCP Server

This project is a small Model Context Protocol (MCP) server that lets you query a DuckDB-based employee database through a simple tool interface. It is a beginner-friendly example for exploring how MCP servers work and how SQL can be exposed to AI assistants or local tools.
If you are new to this project, the best way to start is:
- set up the environment,
- seed the sample database,
- run the MCP server locally,
- try a few SQL queries.
What this project does
The server exposes a single tool named run_sql that accepts a SQL statement and returns the result as a Markdown table. The sample database contains employee, department, salary, and attendance data so you can explore join queries and reporting-style SQL.
Project files
duck_db_mcp.py- the MCP server implementationseed_employee_db.py- creates and populates the DuckDB databaseemployee.duckdb- the generated database filepyproject.toml- project dependenciesMakefile- helper commands for setup
Quick start
1. Create and activate a virtual environment
uv init
uv venv
2. Install dependencies
uv add "mcp[cli]" fastmcp duckdb pandas tabulate
3. Create the sample database
make setup-db
You can also run this directly:
python3 seed_employee_db.py
Run the MCP server locally
Test the server in development mode
mcp dev duck_db_mcp.py
Or with uv:
uv run mcp dev duck_db_mcp.py
Install the MCP server
mcp install duck_db_mcp.py
Start exploring
Once the server is running, try simple queries like these:
Example: list employees and departments
SELECT
e.first_name,
e.last_name,
d.department_name
FROM employee e
JOIN department d
ON e.department_id = d.department_id
ORDER BY e.last_name;
Example: view salary information
SELECT
e.first_name,
e.last_name,
s.net_salary,
s.pay_month
FROM employee e
JOIN salary s
ON e.employee_id = s.employee_id
ORDER BY s.net_salary DESC;
Example: employee attendance summary
SELECT
e.first_name,
e.last_name,
a.attendance_date,
a.status
FROM employee e
JOIN attendance a
ON e.employee_id = a.employee_id
ORDER BY a.attendance_date, e.last_name;
Claude Desktop MCP config
If you want to use this server from Claude Desktop, install it first:
mcp install duck_db_mcp.py
A typical configuration looks like this:
{
"DuckDB SQL Server": {
"command": "/opt/homebrew/bin/uv",
"args": [
"run",
"--frozen",
"--with",
"mcp[cli]",
"mcp",
"run",
"/Users/rahulkumar/Desktop/AI/mcp_server_learning/duck_db_mcp/duck_db_mcp.py"
],
"env": {
"VIRTUAL_ENV": "/Users/rahulkumar/Desktop/AI/mcp_server_learning/duck_db_mcp/.venv",
"PATH": "/Users/rahulkumar/Desktop/AI/mcp_server_learning/duck_db_mcp/.venv/bin:/opt/homebrew/bin:/usr/bin:/bin"
}
}
}
Install from GitHub
If you want someone else to install this package directly from GitHub, use this command:
uvx --from git+https://github.com/panditrahulsharma/duck_db_mcp.git duck_mcp
For Claude Desktop, the config can look like this:
{
"mcpServers": {
"DuckDB SQL Server": {
"command": "uvx",
"args": [
"--from",
"git+https://github.com/panditrahulsharma/duck_db_mcp.git",
"duck_mcp"
]
}
}
}
Make sure the latest changes in pyproject.toml and duck_db_mcp.py have been committed and pushed to GitHub before trying this install method.
Git MCP config
If you also want to use Git-related MCP tools, you can add a Git MCP server configuration in the same client config. A typical example is:
{
"mcpServers": {
"DuckDB SQL Server": {
"command": "uvx",
"args": [
"--from",
"git+https://github.com/panditrahulsharma/duck_db_mcp.git",
"duck_mcp"
]
}
}
}
This is useful when you want your MCP client to access Git operations such as status, diff, branch, and commit history while working with this project.
Notes for contributors
When adding or changing tools, make sure the docstring is clear and descriptive. A well-written docstring helps MCP clients understand what the tool does and how to use it correctly.
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。