Assignment 3: Customer 360 Operator Agent

Objective

Build a Model Context Protocol (MCP) Server that wraps an existing Core Banking Microservice and an MCP Client Chatbot that interacts with it. This assignment focuses on standardizing tool exposure using the MCP specification.

The Existing Service

You are provided with a simulated Core Banking Customer Microservice.

• Base URL: https://customer-microservice-eu.proudsky-2219dea3.westeurope.azurecontainerapps.io
• Interactive Docs: Swagger UI

Capabilities

• Get Customer Profile: GET /api/v1/customers/{id}
• List Accounts: GET /api/v1/customers/{id}/accounts
• Internal Transfer: POST /api/v1/transfers/internal

Test Data

The database is in-memory and resets on restart. Use these IDs:

• Alice Johnson: CUST-001 (Has 2 accounts)
• Bob Smith: CUST-002 (Has 1 account)

Core Tasks

1. Build the MCP Server

• Create a new server using FastMCP
• This server should expose the Core Banking Microservice's POST endpoints as MCP Tools and GET endpoints as MCP Resources
• The tools should handle the API calls to the provided FastAPI app.

2. Build the MCP Client Chatbot

• Build a chatbot application that acts as an MCP Client.
• It should connect to your MCP Server.
• Users should be able to ask natural language questions like:
  • "What is Alice's current balance?"
  • "Transfer $500 from Alice's checking to her savings."
• The chatbot uses the LLM to select the appropriate MCP tool and execute the action.

3. Frontend Reactivity

• The frontend must be reactive to agent processing.
• Visibility: When the agent calls a tool (e.g., transfer_money), the UI should display this action in real-time (e.g., "Agent is executing transfer...").
• Technical Hint: Use Server-Sent Events (SSE) to stream events from the backend to the frontend. The frontend should listen to these events and update the UI accordingly.

Suggested Team Split

To complete this assignment effectively as a team of 3, we recommend the following division of labor:

• Member 1: MCP Server Implementation
  • Analyzing the upstream API.
  • Implementing the FastMCP server.
  • Defining the Tool (POST) and Resource (GET) definitions.
• Member 2: Frontend (Streamlit)
  • Building the Chatbot UI using Streamlit.
  • Implementing the event loop to render SSE updates.
  • Ensuring a smooth, reactive user experience.
• Member 3: MCP Client & Backend
  • Building the MCP Client logic to connect to the server.
  • Handling the LLM interaction loop.
  • Setting up the backend API to stream events to the frontend.