A2A and MCP Protocols with Java and Spring

Integrating AI Capabilities with Spring Boot: A2A and MCP Protocols Code for this article is available here Introduction to A2A Protocol The Agent-to-Agent (A2A) protocol, developed by Google, enables seamless communication between AI agents and services. It provides a standardized way for AI agents to discover, understand, and interact with various service capabilities. By implementing A2A, your Spring Boot applications become AI-ready, allowing AI agents to naturally interact with your services. Understanding MCP (Model Context Protocol) MCP (Model Context Protocol) is Anthropic's specification for enabling structured interactions between AI models and external tools/services. It provides a standardized way to define tool interfaces that can be called by AI models, making services discoverable and executable in a controlled manner. This protocol is particularly useful for creating safe and reliable integrations between AI models and external capabilities. Key Features of MCP: Structured tool definitions with clear input/output schemas Runtime type validation Standardized error handling Support for streaming responses Built-in security considerations Creating AI-Enabled Spring Boot Services Service Implementation Examples The framework supports several types of services that can be exposed to AI agents: CompareCarService: Compares features between two cars @Service @Agent(groupName ="compareCar", groupDescription = "Provide 2 cars and compare them") public class CompareCarService { @Action(description = "compare 2 cars") public String compareCar(String car1, String car2) { return car2 + " is better than " + car1; } } WeatherService: Provides weather information for cities @Service @Agent(groupName = "getTemperature", groupDescription = "get weather for city") public class WeatherService { @Action(description = "get temperature for city") public String getTemperature(String cityName) { // Weather API integration return "Temperature for " + cityName; } } Customer Service Integration: Handle customer support tickets @Service @Agent(groupName ="raiseTicket", groupDescription = "Create a ticket for customer") public class RaiseCustomerTicket { @Action(description = "Raise a ticket for customer") public String raiseTicket(String customerName) { return "ticket raised for " + customerName; } } Protocol Integration A2A Agent Card The A2A agent card describes your service capabilities to AI agents. It includes: Service metadata (name, description, version) Available skills Input/output modes Authentication requirements A2A Agent Card Example When an AI agent discovers your service, it receives a detailed agent card in JSON format that describes all available capabilities. Here's what your service exposes: { "name": "Helpful Agent", "description": "This agent provides capabilities to raise tickets, compare cars, find out a person's favorite food, and get weather information.", "url": "http://localhost:7860", "provider": { "organization": "Example Corp", "url": "https://example.com" }, "version": "1.0", "documentationUrl": "https://example.com/helpfulagent/docs", "capabilities": { "streaming": false, "pushNotifications": false, "stateTransitionHistory": false }, "authentication": { "schemes": ["API Key"], "credentials": "API Key" }, "defaultInputModes": ["Text", "Voice"], "defaultOutputModes": ["Text", "Voice"], "skills": [ { "id": "raiseTicket", "name": "Raise Ticket", "description": "Create a ticket for a customer.", "tags": ["Ticket", "Support", "Customer Service"], "examples": [ "Raise a ticket for John Doe", "Create a support ticket" ], "inputModes": ["Text", "Voice"], "outputModes": ["Text"] }, { "id": "compareCar", "name": "Compare Cars", "description": "Compare two cars based on their features.", "tags": ["Car", "Comparison", "Automobile"], "examples": [ "Compare Toyota Camry and Honda Accord", "Compare two cars" ], "inputModes": ["Text"], "outputModes": ["Text"] }, { "id": "whatThisPersonFavFood", "name": "Find Favorite Food", "description": "Find out what a person likes to eat.", "tags": ["Food", "Preferences", "Person"], "examples": [ "What does John like to eat?", "Find out Mary's favorite food" ], "inputModes": ["Text", "Voice"], "outputModes": ["Text"] }, { "id": "getTemperature", "name": "Get Temperature", "description": "Get the current weather temperature for a city.", "tags": ["Weather", "Temperature", "City"], "examples": [ "What is the weather in London?", "Get the temp

May 8, 2025 - 00:30

A2A and MCP Protocols with Java and Spring

Integrating AI Capabilities with Spring Boot: A2A and MCP Protocols

Code for this article is available here

Introduction to A2A Protocol

The Agent-to-Agent (A2A) protocol, developed by Google, enables seamless communication between AI agents and services. It provides a standardized way for AI agents to discover, understand, and interact with various service capabilities. By implementing A2A, your Spring Boot applications become AI-ready, allowing AI agents to naturally interact with your services.

Understanding MCP (Model Context Protocol)

MCP (Model Context Protocol) is Anthropic's specification for enabling structured interactions between AI models and external tools/services. It provides a standardized way to define tool interfaces that can be called by AI models, making services discoverable and executable in a controlled manner. This protocol is particularly useful for creating safe and reliable integrations between AI models and external capabilities.

Key Features of MCP:

Structured tool definitions with clear input/output schemas
Runtime type validation
Standardized error handling
Support for streaming responses
Built-in security considerations

Creating AI-Enabled Spring Boot Services

Service Implementation Examples

The framework supports several types of services that can be exposed to AI agents:

CompareCarService: Compares features between two cars

@Service
@Agent(groupName ="compareCar", groupDescription = "Provide 2 cars and compare them")
public class CompareCarService {
    @Action(description = "compare 2 cars")
    public String compareCar(String car1, String car2) {
        return car2 + " is better than " + car1;
    }
}

WeatherService: Provides weather information for cities

@Service
@Agent(groupName = "getTemperature", groupDescription = "get weather for city")
public class WeatherService {
    @Action(description = "get temperature for city")
    public String getTemperature(String cityName) {
        // Weather API integration
        return "Temperature for " + cityName;
    }
}

Customer Service Integration: Handle customer support tickets

@Service
@Agent(groupName ="raiseTicket", groupDescription = "Create a ticket for customer")
public class RaiseCustomerTicket {
    @Action(description = "Raise a ticket for customer")
    public String raiseTicket(String customerName) {
        return "ticket raised for " + customerName;
    }
}

Protocol Integration

A2A Agent Card

The A2A agent card describes your service capabilities to AI agents. It includes:

Service metadata (name, description, version)
Available skills
Input/output modes
Authentication requirements

A2A Agent Card Example

When an AI agent discovers your service, it receives a detailed agent card in JSON format that describes all available capabilities. Here's what your service exposes:

{
  "name": "Helpful Agent",
  "description": "This agent provides capabilities to raise tickets, compare cars, find out a person's favorite food, and get weather information.",
  "url": "http://localhost:7860",
  "provider": {
    "organization": "Example Corp",
    "url": "https://example.com"
  },
  "version": "1.0",
  "documentationUrl": "https://example.com/helpfulagent/docs",
  "capabilities": {
    "streaming": false,
    "pushNotifications": false,
    "stateTransitionHistory": false
  },
  "authentication": {
    "schemes": ["API Key"],
    "credentials": "API Key"
  },
  "defaultInputModes": ["Text", "Voice"],
  "defaultOutputModes": ["Text", "Voice"],
  "skills": [
    {
      "id": "raiseTicket",
      "name": "Raise Ticket",
      "description": "Create a ticket for a customer.",
      "tags": ["Ticket", "Support", "Customer Service"],
      "examples": [
        "Raise a ticket for John Doe",
        "Create a support ticket"
      ],
      "inputModes": ["Text", "Voice"],
      "outputModes": ["Text"]
    },
    {
      "id": "compareCar",
      "name": "Compare Cars",
      "description": "Compare two cars based on their features.",
      "tags": ["Car", "Comparison", "Automobile"],
      "examples": [
        "Compare Toyota Camry and Honda Accord",
        "Compare two cars"
      ],
      "inputModes": ["Text"],
      "outputModes": ["Text"]
    },
    {
      "id": "whatThisPersonFavFood",
      "name": "Find Favorite Food",
      "description": "Find out what a person likes to eat.",
      "tags": ["Food", "Preferences", "Person"],
      "examples": [
        "What does John like to eat?",
        "Find out Mary's favorite food"
      ],
      "inputModes": ["Text", "Voice"],
      "outputModes": ["Text"]
    },
    {
      "id": "getTemperature",
      "name": "Get Temperature",
      "description": "Get the current weather temperature for a city.",
      "tags": ["Weather", "Temperature", "City"],
      "examples": [
        "What is the weather in London?",
        "Get the temperature for New York"
      ],
      "inputModes": ["Text", "Voice"],
      "outputModes": ["Text"]
    }
  ]
}

This agent card provides a complete description of your service's capabilities:

Basic Information: Name, description, version, and provider details
Capabilities: Supported features like streaming and notifications
Authentication: Required security mechanisms
Input/Output Modes: Supported interaction methods
Skills: Detailed list of available actions with:
- Unique IDs
- Human-readable names
- Descriptions
- Example usage
- Supported input/output modes
- Relevant tags

MCP Tools Interface

MCP tools are exposed through a dedicated endpoint that provides:

Tool definitions
Parameter schemas
Input validation rules
Response formats

Example MCP tools endpoint: http://localhost:7860/mcp/list-tools

MCP Tools Response Example

Similarly, when accessed through the MCP endpoint, your tools are described in a structured format:

{
  "tools": [
    {
      "parameters": {
        "type": "object",
        "properties": {
          "provideAllValuesInPlainEnglish": {
            "type": "string",
            "description": "{\n    \"customerName\": \"String\"\n}"
          }
        },
        "required": ["provideAllValuesInPlainEnglish"],
        "additionalProperties": false
      },
      "inputSchema": {
        "type": "object",
        "properties": {
          "provideAllValuesInPlainEnglish": {
            "type": "string",
            "description": "{\n    \"customerName\": \"String\"\n}"
          }
        },
        "required": ["provideAllValuesInPlainEnglish"]
      },
      "description": "Raise a ticket for customer",
      "name": "raiseTicket"
    }
    // ... other tools definitions ...
  ]
}

Controller Implementation

A2A Controllers

MyA2ACardController: Manages the agent card endpoint
MyA2ATaskController: Handles task execution
MyRpcController: Processes JSON-RPC requests

MCP Controller

The MCPController handles tool discovery and execution:

@RestController
@RequestMapping("/mcp")
public class MCPController extends MCPToolsController {
    @GetMapping("/list-tools")
    public ResponseEntity<Map<String, List<Tool>>> listTools() {
        // Returns available tools
    }

    @PostMapping("/call-tool")
    public ResponseEntity<JSONRPCResponse> callTool(@RequestBody ToolCallRequest request) {
        // Executes tool calls
    }
}

Making Services AI-Accessible

To expose your Spring services to AI agents:

Add the required dependencies:


    io.github.vishalmysore
    a2ajava
    0.0.7.2


    io.github.vishalmysore
    tools4ai
    1.0.7.5

Annotate your services with @Agent and @Action
Configure the application properties:

spring.application.name=spring-boot
server.port=7860
a2a.persistence=cache

Testing AI Integration

You can test your AI-enabled services through:

Direct HTTP calls to service endpoints
A2A agent card verification
MCP tool discovery and execution
Natural language prompts through the action endpoint

Example test prompt:

GET http://localhost:7860/action?prompt="Compare Honda Civic and Toyota Corolla"

Conclusion

By implementing both A2A and MCP protocols, your Spring Boot applications become fully AI-capable, allowing seamless integration with various AI agents and systems. This dual-protocol support ensures maximum compatibility and flexibility in the AI ecosystem.