A2Aassignment

Multi-Agent Customer Service System

Using MCP, A2A Protocol, and LangGraph

A production-ready multi-agent system demonstrating Agent-to-Agent (A2A) communication protocol with Model Context Protocol (MCP) for database operations. Built with official SDKs and LangGraph for agent orchestration.

🏗️ System Architecture

User Query
    ↓
[A2A Client] (JSON-RPC)
    ↓
[Router Agent] ← Analyzes & coordinates
    ↓
[Data Agent] ← MCP Server ← SQLite Database
    ↓
[Support Agent] ← Generates responses

Components:

• MCP Server: 5 tools for customer/ticket operations
• Data Agent: Executes database operations via MCP
• Support Agent: Provides customer assistance
• Router Agent: Orchestrates multi-agent workflows
• SQLite Database: 30 customers, 60+ tickets

📋 Prerequisites

• Python 3.10 or higher
• pip package manager
• Virtual environment support

🚀 Setup Instructions

1. Create Virtual Environment

# Create virtual environment
python -m venv venv

# Activate virtual environment
# On macOS/Linux:
source venv/bin/activate
# On Windows:
venv\Scripts\activate

2. Install Dependencies

pip install -r requirements.txt

3. Set Up API Key

Create a .env file or set environment variable:

export ANTHROPIC_API_KEY="your-api-key-here"

Or in Google Colab, use Secrets (🔑 icon) to add ANTHROPIC_API_KEY.

4. Initialize Database

python database_setup.py

This creates support.db with sample customer and ticket data.

5. Run the System

python main.py

The system will start three A2A servers on ports 10101-10103.

📦 Requirements

See requirements.txt for complete list:

anthropic>=0.40.0
mcp>=1.0.0
a2a-sdk>=0.1.0
langgraph>=0.2.0
langgraph-cli>=0.1.0
langchain-core>=0.3.0
langchain-anthropic>=0.2.0
starlette>=0.37.0
uvicorn>=0.30.0
httpx>=0.27.0
python-dotenv>=1.0.0
asyncclick>=8.1.0
rich>=13.0.0

🧪 Test Scenarios

The system handles 5 required test scenarios:

1. Simple Query

"Get customer information for ID 5"
→ Data Agent → MCP get_customer → Response

2. Coordinated Query

"I'm customer 5 and need help upgrading my account"
→ Router → Data Agent (fetch) → Support Agent (help)

3. Complex Query

"Show me all active customers who have open tickets"
→ Router → Multiple data operations → Synthesis

4. Escalation

"I've been charged twice, please refund immediately!"
→ Router (detects urgency) → Priority routing

5. Multi-Intent

"Update my email to new@email.com for customer 2 and show my ticket history"
→ Router → Parallel execution → Combined response

🔧 MCP Tools

Five database operations available:

1. get_customer(customer_id) - Retrieve customer details
2. list_customers(status, limit) - List customers with filters
3. update_customer(customer_id, data) - Update customer info
4. create_ticket(customer_id, issue, priority) - Create support ticket
5. get_customer_history(customer_id) - Get ticket history

🌐 A2A Protocol

Agent Cards: Each agent exposes capabilities via .well-known/agent-card.json

Communication: JSON-RPC 2.0 protocol with message/send method

Endpoints:

• Data Agent: http://localhost:10101/a2a
• Support Agent: http://localhost:10102/a2a
• Router Agent: http://localhost:10103/a2a

📊 Database Schema

customers table:

• id, name, email, phone, status, created_at, updated_at

tickets table:

• id, customer_id, issue, status, priority, created_at

✅ Verification

Run the test suite:

python test_scenarios.py

Expected output: 5/5 tests passing

🎯 Key Features

• ✅ Official MCP SDK integration
• ✅ A2A protocol compliance (JSON-RPC)
• ✅ LangGraph state management
• ✅ Multi-agent coordination
• ✅ Agent Cards for capability discovery
• ✅ Async task execution
• ✅ Context sharing between agents

📚 Project Structure

.
├── README.md
├── requirements.txt
├── .env
├── database_setup.py      # Initialize SQLite database
├── mcp_server.py          # MCP tools implementation
├── agents.py              # LangGraph agents (Data, Support, Router)
├── a2a_server.py          # A2A server setup
├── main.py                # Start all servers
├── test_scenarios.py      # Run 5 test scenarios
└── support.db             # SQLite database (generated)

🐛 Troubleshooting

Port already in use:

# Change ports in a2a_server.py (lines with port numbers)

API Key error:

# Verify ANTHROPIC_API_KEY is set correctly
echo $ANTHROPIC_API_KEY

Import errors:

# Ensure virtual environment is activated
# Reinstall dependencies
pip install -r requirements.txt --force-reinstall

📝 Assignment Requirements Met

• ✅ MCP Server with 5 tools
• ✅ Three LangGraph agents with proper state
• ✅ A2A protocol implementation
• ✅ Agent Cards for discovery
• ✅ All 5 test scenarios passing
• ✅ Multi-agent coordination demonstrated

🔗 References

• MCP Documentation
• A2A Protocol Spec
• LangGraph Docs

---

Built with: Python 3.12 | MCP SDK | A2A SDK | LangGraph | Claude Sonnet 4