Database Targeting for Formatted SQL
Last updated: May 15, 2026
When the Liquibase AI Changelog Generator generates Formatted SQL output, it needs to know which database you are targeting. Different databases use different SQL syntax for common operations; the same change can require meaningfully different SQL depending on your platform. The LIQUIBASE_MCP_SQL_DATABASE environment variable allows you to tell the AI Changelog Generator which database to generate SQL for.
Before you begin
This procedure requires LIQUIBASE_MCP_OUTPUT_FORMAT to be set to formatted-sql. If you are using XML output, this configuration is not required. XML output is database-agnostic and does not require a target database.
Identify the value for your target database. You will need this in the steps below.
Database | Value | Notes |
PostgreSQL |
| Default |
MySQL |
| |
MariaDB |
| |
Oracle |
| |
SQL Server |
| |
DB2 |
| |
H2 |
| Useful for local testing and CI |
SQLite |
|
Procedure
Set LIQUIBASE_MCP_SQL_DATABASE to one of the following values:
Set LIQUIBASE_MCP_SQL_DATABASE alongside LIQUIBASE_MCP_OUTPUT_FORMAT in your AI Changelog Generator configuration. The examples below show both variables together.
VS Code users:
Add both variables to your AI Changelog Generator entry in VS Code settings.json.
Replace postgresql with the value for your target database. See Supported Databases above for the full list of values.
VSCode MCP server configuration
Docker users:
Pass both variables using the -e flag in your docker run command. Add one -e flag for each variable, before specifying the image name:
Docker MCP Server configuration
Claude Desktop users:
Add both variables to your AI Changelog Generator entry in the Claude Desktop configuration file:
Configuration file location:
macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
Windows: %APPDATA%\Claude\claude_desktop_config.json
Claude Desktop configuration
Restart the AI Changelog Generator server
Restart required. After changing this variable in VS Code or Claude Desktop, restart the AI Changelog Generator server for the change to take effect.