Spring AI Model Context Protocol Demo Application

A demo application showcasing the integration of Spring AI with File system using the Model Context Protocol (MCP). This application enables natural language interactions with predefiend folders in your local files system.

It starts and connects to Filesystem MCP-Server with provided accsss to your model-context-protocol/filesystem/target folder

Features

Natural language querying and updateing files on your local file system
Predefined question mode for automated database analysis
Seamless integration with OpenAI's language models
Built on Spring AI and Model Context Protocol

Prerequisites

Java 17 or higher
Maven 3.6+
npx package manager
Git
OpenAI API key

Installation

Install npx (Node Package eXecute): first make sure to install npm and then run:
```
npm install -g npx
```

Clone the repository:

git clone https://github.com/spring-projects/spring-ai-examples.git
cd model-context-protocol/filesystem

Set up your OpenAI API key:

export OPENAI_API_KEY='your-api-key-here'

Build the demo:
```
./mvnw clean install
```

Create a sample text file to explore

Create a sample spring-ai-mcp-overview.txt file under your filesystem/target directory manually or use the create-text-file.sh script

Running the Application

Predefined Questions

Runs through a set of preset questions:

./mvnw spring-boot:run

Architecture Overview

Spring AI's integration with MCP follows a simple chain of components:

MCP Client provides the base communication layer with your filesystem
Function Callbacks expose filesystem operations as AI-callable functions
Chat Client connects these functions to the AI model

The bean definitions are described below, starting with the ChatClient

Chat Client

@Bean
@Profile("!chat")
public CommandLineRunner predefinedQuestions(ChatClient.Builder chatClientBuilder,
                                           List<McpFunctionCallback> functionCallbacks,
                                           ConfigurableApplicationContext context) {
    return args -> {
        var chatClient = chatClientBuilder.defaultFunctions(functionCallbacks)
                .build();
        // Run Predefined Questions
    };
}

The chat client setup is remarkably simple - it just needs the function callbacks that were automatically created from the MCP tools. Spring's dependency injection handles all the wiring, making the integration seamless.

Now let's look at the other bean definitions in detail...

Function Callbacks

The application registers MCP tools with Spring AI using function callbacks:

@Bean
public List<McpFunctionCallback> functionCallbacks(McpSyncClient mcpClient) {
    return mcpClient.listTools(null)
            .tools()
            .stream()
            .map(tool -> new McpFunctionCallback(mcpClient, tool))
            .toList();
}

Purpose

This bean is responsible for:

Discovering available MCP tools from the client
Converting each tool into a Spring AI function callback
Making these callbacks available for use with the ChatClient

How It Works

mcpClient.listTools(null) queries the MCP server for all available tools
- The null parameter represents a pagination cursor
- When null, returns the first page of results
- A cursor string can be provided to get results after that position
.tools() extracts the tool list from the response
Each tool is transformed into a McpFunctionCallback using .map()
These callbacks are collected into an array using .toList()

Usage

The registered callbacks enable the ChatClient to:

Access MCP tools during conversations
Handle function calls requested by the AI model
Execute tools against the MCP server (e.g., filesystem)

MCP Client

The application uses a synchronous MCP client to communicate with the Filesystem MCP Server running locally:

@Bean(destroyMethod = "close")
public McpSyncClient mcpClient() {
    var stdioParams = ServerParameters.builder("npx")
            .args("-y", "@modelcontextprotocol/server-filesystem", getDbPath())
            .build();

    var mcpClient = McpClient.sync(new StdioServerTransport(stdioParams),
            Duration.ofSeconds(10), new ObjectMapper());

    var init = mcpClient.initialize();
    System.out.println("MCP Initialized: " + init);

    return mcpClient;
}

This configuration:

Creates a stdio-based transport layer that communicates with the npx MCP server
Specifies the location of folders to be used by the filesystem server.
Sets a 10-second timeout for operations
Uses Jackson for JSON serialization
Initializes the connection to the MCP server

The destroyMethod = "close" annotation ensures proper cleanup when the application shuts down.