Installing
These pages set up Liferay Legacy MCP 1.1.0 against an existing Liferay DXP: first the Liferay side, then the server for the people or agents who will use it.
Liferay Legacy MCP is a community app for experimental use, not supported by Liferay. Install it on a test or development copy of your Liferay, not in production.
Before You Begin
Check that you have:
- A Liferay DXP Enterprise Subscription, for Liferay DXP 7.4 U1–U92, or 2024.Q1 to 2025.Q4. Liferay Legacy MCP is experimental and not supported by Liferay, but it runs only with Liferay DXP. See Compatibility for which connection each release supports.
- An administrator account on that Liferay, to create a user, roles and an OAuth 2 application.
- Docker, or another way to run containers, where the server will run: the person’s computer, or a server next to Liferay.
- A test or development copy of your Liferay. Don’t install it in production.
What You’ll Install
| Piece | Where | What It Is |
|---|---|---|
| A service user and two roles | Liferay | The Liferay user the server acts as, with least privilege. |
| An OAuth 2 client extension | Liferay | liferay-mcp-oauth.zip: the server’s connection to Liferay, an OAuth 2 application limited to the scopes the tools need. |
| The server | A container | ghcr.io/liferay-community-apac/liferay-legacy-mcp, in two variants (below). |
| A sign-in application (optional) | Liferay | Only for people who sign in over HTTPS. |
The Images
| Tag | Size | What It Includes |
|---|---|---|
1.1.0-slim |
about 350 MB | Every tool except the page inspection tools, which show as unavailable. |
1.1.0-full |
about 1.2 GB | Also inspect_page and take_screenshot, with a headless Chromium. |
Both run on linux/amd64 and linux/arm64, as a non-root user, write only to /data, and contain no credentials. The package is public:
docker pull ghcr.io/liferay-community-apac/liferay-legacy-mcp:1.1.0-slim
To pin an exact build, use its digest:
| Tag | Digest |
|---|---|
1.1.0-slim |
sha256:6c8f4457d3f7ad79c4d3f2a2f79e72449a9838c839dcf327370d0a3ff0316c13 |
1.1.0-full |
sha256:ed68f831c2a7be57c06520fda711c04982f702d76702ee91a95ce8db4c4e7195 |
Steps
1. Create the Service User
Follow Creating the Service User, on every platform.
2. Connect Liferay, Where It Runs
Deploy the OAuth 2 connection, and get its credentials for the server:
Liferay Self-Hosted (Tomcat)Copy the client extension zip into Liferay, and run the server on the Liferay machine or people's computers.
Liferay Self-Hosted (Docker)Copy the zip into Liferay's container, and run the server in a container that shares Liferay's network.
Liferay PaaSAdd the client extension to your PaaS project, and copy its credentials for the server.
3. Connect the AI Assistant
With Claude CodeAdd the server with one
claude mcp add command, in a terminal.
With the Claude Desktop AppUse the server in the desktop app's chats and its Code tab.
Other AI AssistantsDesktop AI assistants and IDEs that read an mcpServers setting.
For Services and PeopleOne shared server over HTTPS, with API keys for services and Liferay sign-in for people.
Then follow Checking the Installation.