dev-model-context-protocol

📦 Container Booking POC with Agent Orchestration and MCP Server

This repository presents a proof of concept (PoC) for orchestrating a multi-step container booking workflow using the Model Context Protocol (MCP), built with .NET and Semantic Kernel.

🧭 Workflow Overview

The container booking process is broken into three distinct steps, each handled by a specialized agent:

1. Find Vessel – Searches available vessels based on user input
2. Check Capacity – Checks space availability on the selected vessel
3. Complete Booking – Finalizes the booking process with validated details

Each agent connects to a separate MCP server, enabling decentralized knowledge and function isolation. An orchestrator interprets user intent using natural language and invokes the appropriate agent accordingly.

🔐 Security

All MCP servers are secured using Microsoft Entra ID (formerly Azure Active Directory).

• Authentication: Agents authenticate using OAuth2 client credentials
• Authorization: Role-based access control (RBAC) ensures agents only access permitted scopes

🧠 Tech Stack

• C# (.NET) – Core implementation of agents, orchestration logic, and MCP clients
• Azure OpenAI Service – For natural language understanding and orchestration logic
• GPT-4 – Model used via Semantic Kernel for interpreting user prompts and selecting agents
• Semantic Kernel – Enables intelligent planning and multi-agent coordination
• Microsoft Entra ID (Azure AD) – Secure access and identity management for MCP servers

🔐 Azure AD Setup for MCP Authentication (Client Credentials Flow)

🧩 Step-by-Step Setup in Azure

🔹 Step 1: Register an Application (Client App)

1. Go to https://portal.azure.com
2. Navigate to Microsoft Entra ID > App registrations
3. Click + New registration
4. Fill out the form:
   • Name: mcp-client-app
   • Supported account types: Single tenant (or as per your needs)
   • Leave Redirect URI empty for now
5. Click Register

---

🔹 Step 2: Generate Client Secret

1. In the registered app, go to Certificates & secrets
2. Click + New client secret
3. Provide a name and expiry
4. Click Add
5. Copy the secret value immediately — you won't be able to see it again

---

🔹 Step 3: Note Down App Info

Field	Source
Tenant ID	From the app's Overview tab
Client ID	Also from the Overview tab
Client Secret	From Step 2
Token Endpoint	https://login.microsoftonline.com/{tenant-id}/oauth2/v2.0/token

---

🔹 Step 4: (Optional but Recommended) Create a Scope for Your API

If you have registered your MCP Server as another Azure AD application (API app), follow these steps:

1. Register your API App

   • Example name: mcp-server-api

2. Expose an API

   • Go to the app > Expose an API
   • Set an Application ID URI (e.g. api://<api-client-id>)

3. Add a Scope

   • Click + Add a scope
   • Fill in the form:
     • Scope name: mcp.readwrite
     • Who can consent: Admins only (or as needed)
     • Admin consent display name: Full MCP access
     • Admin consent description: Allows client apps to read/write to the MCP server.
     • State: Enabled
   • Click Add scope

4. Assign the Scope to the Client App

   • Go to your client app (e.g. mcp-client-app)
   • Navigate to API permissions
   • Click + Add a permission
   • Choose My APIs
   • Select your API app
   • Select the scope you created (mcp.readwrite)
   • Click Add permissions
   • Click Grant admin consent if required

---

🛠️ Step 5: Request a Token Using Client Credentials Flow

Now that your client app is registered and scoped, request a token using the OAuth 2.0 client credentials flow.

You can do this using curl, Postman, or your preferred HTTP client:

curl -X POST https://login.microsoftonline.com/<tenant-id>/oauth2/v2.0/token \
  -H "Content-Type: application/x-www-form-urlencoded" \
  -d "client_id=<client-id>" \
  -d "client_secret=<client-secret>" \
  -d "scope=api://<api-client-id>/.default" \
  -d "grant_type=client_credentials"