Model Context Protocol (MCP) Client Integration for CrafterCMS

Overview

This plugin is a SpringAI-based chat service configured with an MCP (Model Context Protocol) client. This serves as a foundation for an AI-enabled content application, orchestrating Large Language Models while securely exposing content, services, and tools through MCP.

• Built on Spring AI’s orchestration framework
• Uses OpenAI as the default LLM provider
• Designed to be vendor-agnostic (other LLM providers can be swapped in via Spring AI with minimal changes)
• Demonstrates end-to-end MCP integration from UI → LLM → MCP server

Quick Links

Installation Video & Demo https://www.youtube.com/watch?v=iyiW3MgEXzU

---

What MCP Enables

The Model Context Protocol (MCP) provides a standardized way for LLMs to:

• Discover available tools, resources, and capabilities
• Request structured data and perform actions through well-defined interfaces
• Operate with explicit, auditable context rather than implicit prompt stuffing

Within CrafterCMS, MCP enables:

• Secure access to CMS content, metadata, and services

• Controlled execution of server-side tools

• Clear separation between:

  • LLM reasoning
  • Context exposure
  • Application logic

This architecture makes AI integrations safer, more maintainable, and more extensible than traditional prompt-only approaches.

---

Architecture Summary

At a high level, the plugin consists of:

• An MCP client service that exposes CrafterCMS context to the LLM

• A Spring AI chat client responsible for:

  • Prompt orchestration
  • Tool invocation
  • LLM provider abstraction

• A FreeMarker-based chat UI embedded in CrafterCMS

---

Installation & Configuration

Prerequisites

• A valid OpenAI API key

Steps

1. Configure the following environment variable with your OpenAI API key:

   crafter_openai
   

2. Install the plugin into your CrafterCMS project.

Optional parameters (shown here with defaults):

<configuration>
    <mcpServerUrl>/api/plugins/org/craftercms/rd/plugin/mcp/server/craftermcp/mcp.json</mcpServerUrl>
    <aiSystemPrompt>You are a helpful assistant who answers questions in a professional manner.</aiSystemPrompt>
    <aiModelId>o4-mini-2025-04-16</aiModelId>
</configuration>

3. To enable the example chat UI, include the provided FreeMarker template:

   <#include '/templates/plugins/org/craftercms/rd/plugin/mcp/client/ai-chat.ftl' />

Once included, the chat interface will be available within your site and connected to the LLM through Spring AI and MCP.