Liferay Community APAC
Liferay Legacy MCP docs

Home/Marketplace Apps/Liferay Legacy MCP

Getting Started

Liferay Legacy MCP is a server that speaks the Model Context Protocol (MCP), the standard way AI assistants call tools. Connected to your Liferay, it lets an AI assistant work with your sites and content. This page explains how it fits together, and the shortest way to try it.

A community app, for experimental use. Liferay Legacy MCP is shared by Liferay Community APAC. It isn’t a Liferay product: Liferay doesn’t support it, and it comes as is, with no warranty and no support commitment. Use it to explore and evaluate on a test or development copy of your Liferay, not in production. AI assistants can misread a request, so start with writes off, and check every preview before you confirm it. Running it still requires a Liferay DXP Enterprise Subscription.

  1. How It Works
  2. Try It
  3. Related Information

How It Works

The server runs next to Liferay, as a container, and calls Liferay’s APIs:

  • As one service user. A Liferay user you create, with a site role that lets it view content and add and edit what the write tools write, but not delete or change permissions.
  • Through an OAuth 2 application. A Headless Server application that acts as the service user, with only the scopes the tools need. Liferay gives the server its credentials as routes, a folder with one file per value, so no password goes in the server’s settings.

An AI assistant then reaches the server in one of two ways:

Who How Page
One person, in Claude Code, the Claude desktop app, or another AI assistant The assistant starts the server on their computer (stdio) With Claude Code, With the Claude Desktop App, Other AI Assistants
Services, and people who sign in with Liferay Over HTTPS, with API keys or Liferay sign-in For Services and People

Try It

The quickest way to see it work is one person, with a desktop AI assistant, on a test copy of your Liferay:

  1. Create the service user.
  2. Deploy the OAuth 2 client extension, and copy its credentials, as on Tomcat, Docker or PaaS.
  3. Add the server to Claude Code, or to another AI assistant.
  4. Ask: “What Liferay server are you connected to?” The assistant calls get_server_info, which names your Liferay’s release, the connection, and which tools your Liferay supports.

Then try a few questions from What to Ask, such as “Find articles in my-site that haven’t been modified in six months.”

Writes are off until you turn them on. Start read-only, and turn on writes once you’ve seen what the assistant does. See Writes and Previews.