Spring AI MCP Demos

This is a whole MCP stack written in Java with Spring Boot 4.0.x and Spring AI. It implements several demos for the MCP server, MCP client and an Angular chatbot using deep-chat.

Prepare your LLM

Edit the mcp-client/src/main/resources/application.yaml file and set your OpenAI settings. If you want to use a local model, you could also configure Ollama.

Starting the whole Stack

Two convenience scripts are provided at the repository root to start all four services at once: the OAuth2 authorization server, the MCP server, the MCP client and the Angular chatbot.

• Windows: start-all.bat
  • Opens one console window per service (auth server, MCP server, MCP client, chatbot). Close a window (or press Ctrl+C in it) to stop that service.
• Linux/macOS: ./start-all.sh
  • Runs all services as background jobs of the script and writes their output to logs/*.log. Press Ctrl+C to stop everything.

Both scripts start the services in this order, with short delays so each service is ready before the next one that depends on it starts:

1. MCP Authorization Server — http://localhost:9000
2. MCP Server — http://localhost:8082
3. MCP Client — http://localhost:8083
4. MCP Chatbot (Angular) — http://localhost:4200

How to use the Chatbot

1. Open http://localhost:4200 in a browser.
2. Click Sign in and log in with the demo user:
   • Username: john@doe.com
   • Password: john
3. After a successful login you are redirected back to the app, the toolbar shows John Doe, and the deep-chat window connects to the mcp-client so you can start chatting.
4. Ask "Which movie is Kai's favorite quote from?"
5. Approve the tool call in the Chatbot
6. The chatbot will call the MCP client, which calls the MCP server, which calls the OpenAI API to get the answer.
7. Try "What's the weather in Munich?" to see mcp-client ask for your preferred temperature unit before calling the tool, or "What's the weather in Munich in Fahrenheit?" to see it skip that question when you already specified a unit.

Licenses

• mcp-authorization-server is licensed under the Apache License, Version 2.0
• mcp-client is licensed under the Apache License, Version 2.0
• mcp-server is licensed under the Apache License, Version 2.0
• mcp-chatbot is licensed under the MIT License