Model Context Protocol Server for JDBC

This Model Context Protocol(MCP) server enables Large Language Models (LLMs) to inspect, query, create and modify database content given just a Java Database Connectivity (JDBC) url.

Supported databases include PostgresSQL, Oracle, MariaDB, Sqlite and many more.

See video here.

General Usage

Install jbang
Configure your MCP Client to run the server (see Claude Desktop Config below)

Below are examples of command lines to use for configuring the server.

Start with an empty database (in-memory h2 database):

jbang mcp-server-jdbc@quarkusio

A live database (any common JDBC compatible database):

jbang mcp-server-jdbc@quarkusio [jdbc_url]

For example, to connect to a already running Oracle database:

jbang mcp-server-jdbc@quarkusio jdbc:oracle:thin:@localhost:1521:orcl

You can also specify a user and password separately, here for a PostgreSQL database:

jbang mcp-server-jdbc@quarkusio jdbc:postgresql://localhost:5432/sakila -u sakila -p p_ssW0rd

Downloadable databases

JBang can download files from the web and feed them directly to databases like h2 and sqlite.

That is done using the %{url} syntax.

Following is an example of how to connect to a Chinook database in SQLite, downloading the Chinook_Sqlite.sqlite file from github.

jbang mcp-server-jdbc@quarkusio jdbc:sqlite:%{https://github.com/lerocha/chinook-database/raw/master/ChinookDatabase/DataSources/Chinook_Sqlite.sqlite}

There are interesting examples of sample data available online you can use to explore data.

Below are a non-exhaustive list of databases and example JDBC urls that are known to work with this server as-is (no additional configuration required).

Example databases

Chinook

Sample music store data including artists, albums, tracks, invoices etc.

jdbc:sqlite:%{https://github.com/lerocha/chinook-database/raw/master/ChinookDatabase/DataSources/Chinook_Sqlite.sqlite}

Northwind

Classic Microsoft sample database with customers, orders, products etc.

jdbc:sqlite:%{https://github.com/jpwhite3/northwind-SQLite3/raw/refs/heads/main/dist/northwind.db}

Netflix Movies

Sample movie data based on Netflix catalog

jdbc:sqlite:%{https://github.com/lerocha/netflixdb/releases/download/v1.0.0/netflixdb.sqlite}

Supported JDBC variants

Database	Dependency
MariaDB	org.mariadb.jdbc:mariadb-java-client:RELEASE
MySQL	mysql:mysql-connector-java:RELEASE
PostgreSQL	org.postgresql:postgresql:RELEASE
Oracle	com.oracle.database.jdbc:ojdbc10:RELEASE
SQL Server	com.microsoft.sqlserver:mssql-jdbc:RELEASE
SAP DB	com.sapcloud.db.jdbc:ngdbc:RELEASE
Informix	com.ibm.informix:jdbc:RELEASE
Firebird	org.firebirdsql.jdbc:jaybird:RELEASE
hsqldb	org.hsqldb:hsqldb:RELEASE
h2	com.h2database:h2:RELEASE
Apache Derby	org.apache.derby:derby:RELEASE
SQLite	org.xerial:sqlite-jdbc:RELEASE

Components

Below are the MCP components provided by this server.

Tools

read_query - do a SELECT query on the database
write_query - do a INSERT, UPDATE, DELETE query on the database
create_table - create a table in the database
list_tables - list all tables in the database
describe_table - describe a table

Prompts

jdbc_demo - example prompt to get started exploring the server (similar to the one in sqllite MCP)

Claude Desktop Config and mcp-cli

Add this to your claude_desktop.json or server_config.json file:

{
  "mcpServers": {
    "jdbc": {
      "command": "jbang",
      "args": [
        "mcp-server-jdbc@quarkusio",
        "[jdbc_url]"
      ]
    }
  }
}

Troubleshooting

jbang not found

Make sure you have jbang installed and available in your PATH
Alternatively, use full path to jbang executable (e.g. /Users/username/.jbang/jbang)

Get more logging

To get more detailed logging you can add the following parameters to the jbang command line:

-Dquarkus.log.file.enable=true -Dquarkus.log.file.path=${user.home}/mcp-server-jdbc.log

Example:

jbang -Dquarkus.log.file.enable=true -Dquarkus.log.file.path=${user.home}/mcp-server-jdbc.log mcp-server-jdbc@quarkusio jdbc:h2:mem:testdb

Native Image/Executable

Although the server can be compiled to a native executable, it is not recommended as then it is not possible to dynamically load the relevant JDBC driver.

You can however compile a server that targets a specific JDBC driver, but that is for now left up as an exercise to the reader. If you are interested in this, please open issue or PR.

How was this made?

The MCP server uses Quarkus, the Supersonic Subatomic Java Framework and its Model Context Protocol support.

If you want to learn more about Quarkus MCP Server support, please see this blog post and the Quarkus MCP Server extension documentation.

To launch the server and handle the magic setup of JDBC urls it uses jbang to setup Java and run the .jar as transparent as possible. Very similar to how uvx, pipx, npmx and others works; just for Java.