An Enterprise Operations Assistant built with Model Context Protocol (MCP), Java 25 and Spring Boot 3.5.

This project integrates:

• Model Context Protocol (MCP): for communication with external tools via McpSyncClient
• Spring AI 1.1.2: for OpenAI GPT-4 integration
• Spring Boot 3.5: for enterprise architecture

The application retrieves system metrics and recent incidents via MCP, processes them and provides intelligent operational recommendations using GPT-4.

• JDK 25 (or JDK 21+)
• Maven 3.8+
• MCP Server running at http://localhost:8081/mcp (Streamable HTTP)
• OpenAI API Key

Windows (PowerShell):

$env:OPENAI_API_KEY="sk-your-api-key-here"

Linux/Mac:

export OPENAI_API_KEY="sk-your-api-key-here"

Or edit src/main/resources/application.properties and replace ${OPENAI_API_KEY} with your key directly.

The client expects a Streamable HTTP MCP server available at:

http://localhost:8081/mcp

The server must expose the following tools:

• getSystemMetrics → returns { avgLatencyMs, errorRate }
• getRecentIncidents → returns a list of [{ id, title, timestamp, status }]

spring.application.name=operations-mcp-client

# MCP Client
spring.ai.mcp.client.type=SYNC
spring.ai.mcp.client.request-timeout=30s
spring.ai.mcp.client.toolcallback.enabled=false
spring.ai.mcp.client.streamable-http.connections.tools-server.url=http://localhost:8081
spring.ai.mcp.client.streamable-http.connections.tools-server.endpoint=/mcp

# OpenAI
spring.ai.openai.api-key=${OPENAI_API_KEY}
spring.ai.openai.chat.options.model=gpt-4
spring.ai.openai.chat.options.temperature=0.7

mvn clean install

mvn spring-boot:run

curl http://localhost:8080/api/incidents/analyze

Or open in browser: http://localhost:8080/api/incidents/analyze

operations-mcp-client/
├── src/main/java/it/matteoroxis/operations_mcp_client/
│   ├── config/
│   │   ├── McpClientConfig.java           # McpSyncClient configuration
│   │   └── SpringAIConfig.java            # Spring AI ChatClient configuration
│   ├── controller/
│   │   └── IncidentController.java        # REST controller GET /api/incidents/analyze
│   ├── model/
│   │   ├── Incident.java                  # Record: id, title, timestamp, status
│   │   └── SystemMetrics.java             # Record: avgLatencyMs, errorRate, timestamp
│   ├── service/
│   │   └── IncidentAnalysisService.java   # Core analysis service
│   └── OperationsMcpClientApplication.java
└── src/main/resources/
    └── application.properties

GET /api/incidents/analyze
        │
        ▼
IncidentAnalysisService.analyze()
        │
        ├── McpSyncClient → callTool("getSystemMetrics")   → SystemMetrics
        ├── McpSyncClient → callTool("getRecentIncidents") → List<Incident>
        │
        ▼
ChatClient (OpenAI GPT-4)
        │
        ▼
Operational recommendations in natural language

Incident

public record Incident(String id, String title, Instant timestamp, String status) {}

SystemMetrics

public record SystemMetrics(int avgLatencyMs, double errorRate, Instant timestamp) {}

The service dynamically builds a prompt containing:

• System metrics: average latency, error rate, timestamp
• Recent incidents list: ID, title, status, timestamp
• Analysis constraints: likely causes, prioritised actions, additional metrics to collect

The MCP server is not available at http://localhost:8081. Make sure the server is running before starting the client.

Maven cannot find the JDK. Make sure JAVA_HOME points to a JDK:

# Windows
$env:JAVA_HOME="C:\Program Files\Java\jdk-25"

Verify that the OPENAI_API_KEY environment variable is set correctly.

Make sure the MCP server returns JSON in the expected format:

[
  { "id": "INC-42", "title": "...", "timestamp": 1234567890, "status": "Resolved" }
]

The timestamp field must be an epoch number in seconds (not an ISO-8601 string).

• Java 25
• Spring Boot 3.5.10
• Spring AI 1.1.2
• MCP Java SDK (via spring-ai-starter-mcp-client)
• OpenAI GPT-4
• Jackson
• Maven

This project is an educational example.

Matteo Roxis