Configure the Liquibase AI Changelog Generator in VS Code
Last updated: May 15, 2026
Both Liquibase XML changesets and formatted SQL come with their own complexity: XML requires precise syntax and schema expertise, while formatted SQL demands learning specific rules and conventions. The Liquibase AI Changelog Generator solves this by letting you describe database changes in natural language and automatically converting them into valid, tested Liquibase XML or SQL. This Model Context Protocol (MCP) server runs locally and integrates seamlessly with Claude Desktop, turning conversational requests like "add an email column to the users table" into production-ready changesets.
Powered by Liquibase's native Java APIs, the server offers nineteen specialized tools that handle everything from basic table operations to complex procedures and sequences. Each generated changeset is automatically validated against an H2 database to catch errors before they reach your production environment, with configurable validation and rollback options to match your workflow. As a standalone executable with 322 comprehensive tests, you can generate validated changesets immediately with no external Liquibase CLI installation required.
Before you begin
Ensure Liquibase Secure is installed.
Download the JAR file from the email sent to you. You will need to authenticate to access the JAR file.
VS Code 1.102 or later must be installed.
Java 17 or higher is installed and available in the PATH.
An AI agent that can interact with an AI Changelog Generator, such as GitHub CoPilot or Claude Code must be installed.
Procedure
Open VS Code and ensure that CoPilot is enabled.
Open the Command Palette.
Use one of these commands depending on your OS:
Windows/Linux:
Ctrl+Shift+PMac:
Cmd+Shift+P
Configure the AI Changelog Generator via the Command Palette.
Type in: MCP: Add Server
Type in STDIO for the Server Type.
STDIO stands for Standard Input/Output. This is the most common and reliable server transport method.
Enter this command in the command palette to run the server jar file.
Enter java -jar /full/path/to/liquibase-mcp-changelog-server-0.1.0.jar to run the server.
Note: Ensure you replace /full/path/to/ your local path to the jar file downloaded earlier.
Give the server a unique name that indicates it's for Liquibase changelogs.
In this example, we use liquibase-changelog-mcp-server.
Choose your configuration scope.
Select Global to use the AI Changelog Generator across all projects.
Select Workspace to use the AI Changelog Generator only in the current project.
Optional: Configure Formatted SQL output and database targeting
By default, the AI Changelog Generator generates XML output, which is database-agnostic and works with all Liquibase features. If your team prefers SQL-first workflows, you can switch to Formatted SQL output and specify your target database dialect.
Open your settings.json file and add environment variables to the server entry. The following example configures the server to generate Formatted SQL output targeting a PostgreSQL database:
Generate formatted SQL output targeting a PostgreSQL database:
Environment variable reference
Variable | Values | Default | Description |
|
|
| Controls the output format of generated changesets. |
|
|
| Specifies the target database dialect for Formatted SQL generation. Only applies when |
For a full explanation of both formats and guidance on which to choose, see Output Formats in the Liquibase AI Changelog Generator.
For details on database-specific SQL differences, see Database Targeting for Formatted SQL.
Verify the server is installed successfully.
Once complete, the mcp.json file will open to display your server data. Ensure that the server displayed says it is running.
OR:
Open the Command Palette and type in MCP: List Servers
Confirm that the liquibase-changelog-mcp-server appears and says Running in the list.
