LABSCO SUMMARY

Writes are off until you turn them on.

That default is the reason to prefer this over the upstream image, and it pairs naturally with a read-only database user — belt and braces on the same connection. The one setup detail that catches people is stdio mode: without the console-logging switches, Quarkus writes to the same stream the protocol uses.

What it is

A fork of the Quarkus MCP JDBC server that adds two things: a published Docker image so you do not build anything, and a safety default where INSERT, UPDATE and DELETE are disabled until you switch them on.

What you get

Database access over JDBC for an assistant — inspecting, querying, and, when enabled, modifying content
Any database with a JDBC driver the upstream project supports, selected purely by the connection URL
Write operations off by default behind a single enable.write.sql setting
Two transports from the same image: an SSE endpoint at /mcp/sse, or stdio for local clients
Configuration entirely through environment variables, so credentials come from an env file rather than a config file in the repo
Docker Compose examples for both the SSE service and the stdio client setup

Requirements

Docker, and a reachable database with its JDBC connection URL. Set jdbc.url, and jdbc.user and jdbc.password where the database needs them; leave enable.write.sql false unless you mean it. For SSE, publish port 8080 and connect to /mcp/sse. For stdio you must also pass quarkus.mcp.server.stdio.enabled=true together with the two console-logging settings that silence stdout, and run the container with -i — otherwise log output corrupts the protocol stream. Apache licensed.

Setup effort

One command plus credentials — docker run --rm -e enable.write.sql=false -e jdbc.user=db_user -e jdbc.password=db_password -e jdbc.url=jdbc:postgresql://host:port/database -e quarkus.mcp.server.stdio.enabled=true -e quarkus.log.console.enable=false -e quarkus.log.console.stderr=false -i guang1/jdbc-mcp:latest, then supply credentials

See installation options

USE CASES

01

Letting an assistant answer questions from a real database

Point the container at the JDBC URL with writes disabled, and every answer comes from a live read.

02

Exploring a schema you inherited

Ask what tables and columns exist and query them, without writing SQL against a database you do not know yet.

03

Allowing changes deliberately, for one session

Enable write SQL only on the run where you need it, and turn it back off when that work is done.

BEFORE YOU INSTALL

What to know first

Requires

docker on PATH

Where it runs

Both documented setups run the server on your machine.

INSTALLATION

Connect your client

Server setup

TERMINAL

docker run --rm -e enable.write.sql=false -e jdbc.user=db_user -e jdbc.password=db_password -e jdbc.url=jdbc:postgresql://host:port/database -e quarkus.mcp.server.stdio.enabled=true -e quarkus.log.console.enable=false -e quarkus.log.console.stderr=false -i guang1/jdbc-mcp:latest

MCP Clients

Claude Desktop

CLAUDE DESKTOP · JSON

{
  "mcpServers": {
    "jdbc": {
      "command": "docker",
      "args": [
        "run",
        "--rm",
        "-e",
        "enable.write.sql=false",
        "-e",
        "jdbc.user=db_user",
        "-e",
        "jdbc.password=db_password",
        "-e",
        "jdbc.url=jdbc:postgresql://host:port/database",
        "-e",
        "quarkus.mcp.server.stdio.enabled=true",
        "-e",
        "quarkus.log.console.enable=false",
        "-e",
        "quarkus.log.console.stderr=false",
        "-i",
        "guang1/jdbc-mcp:latest"
      ]
    }
  }
}

Claude Desktop — env file

CLAUDE DESKTOP — ENV FILE · JSON

{
  "mcpServers": {
    "jdbc": {
      "command": "docker",
      "args": [
        "run",
        "--rm",
        "--env-file",
        ".env",
        "-e",
        "quarkus.mcp.server.stdio.enabled=true",
        "-e",
        "quarkus.log.console.enable=false",
        "-e",
        "quarkus.log.console.stderr=false",
        "-i",
        "guang1/jdbc-mcp:latest"
      ]
    }
  }
}

Docker reads --env-file .env relative to the folder it is started from, and Claude Desktop does not start it in yours, so .env has to be the full path to the file. The file holds the enable.write.sql, jdbc.user, jdbc.password and jdbc.url lines shown in the project README.

JDBC-MCP