MCP Server Prototype

A prototype of creating MCP server

Model Context Protocol (MCP) Server

What is MCP?

The Model Context Protocol (MCP) is an open standard developed by Anthropic that defines how AI models communicate with external tools, data sources, and services. It acts as a universal connector — similar to how USB-C standardizes device connections — allowing AI agents to discover and invoke capabilities exposed by any MCP-compliant server.

How It Works

┌─────────────┐        MCP Protocol        ┌─────────────┐
│  AI Client  │ ◄────────────────────────► │  MCP Server │
│ (Claude,    │   Tools / Resources /      │  (This app) │
│  Cursor...) │        Prompts             └─────────────┘
└─────────────┘

MCP Features

MCP Server

Primitives

An MCP server exposes three types of primitives:

Primitive	Explaination	Example	Who controls it
Tools	Functions that your LLM can actively call, and decides when to use them based on user requests. Tools can write to databases, call external APIs, modify files, or trigger other logic.	Search, send email, query DB	Model
Resources	Passive data sources that provide read-only access to information for context, such as file contents, database schemas, or API documentation.	Files, API responses, Read calendars	Application
Prompts	Pre-built instruction templates that tell the model to work with specific tools and resources.	Summarise, translate, draft an email	User

Transports

MCP supports multiple transport types:

Transport	Use Case
Stdio	Local processes, CLI tools
Streamable HTTP	Remote/hosted servers (recommended)

MCP Client

Feature	Explaination	Example
Elicitation	Elicitation enables servers to request specific information from users during interactions, providing a structured way for servers to gather information on demand.	User's preferences e.g preferred contact number, confirmation before booking
Roots	Roots allow clients to specify which directories servers should focus on, communicating intended scope through a coordination mechanism.	Access to specifice file directory e.g. specific sharepoint folder
Sampling	Sampling allows servers to request LLM completions through the client, enabling an agentic workflow. This approach puts the client in complete control of user permissions and security measures.	Send list of schools to an LLM to choose best school

API Reference

POST /mcp

Body: JSON-RPC 2.0 object.

Supported methods:

Method	Description
initialize	Start session, negotiate protocol
initialized	Client acknowledgement
ping	Heartbeat
tools/list	List scope-filtered tools
tools/call	Execute a tool
roots/list	List available roots
sampling/createMessage	Request LLM completion from client
elicitation/create	Request user input from client

Getting Started

Prerequisites

• .NET 10 SDK
• Node.js (for MCP Inspector)

Configuration

Add the following to your secrets.json (User secrets) and update with relevant information:

{
  "AzureSearch:Endpoint": "https://<your-resource>.search.windows.net",
  "AzureSearch:ApiKey": "[API-Key]",
  "AzureSearch:Indexes:Ofsted": "[Ofsted Index]",
  "AzureSearch:Indexes:Establishment": "[Establishment Index]"
}

Run the Server

dotnet run

The MCP endpoint will be available at:

http://localhost:7146/mcp

Testing with MCP Inspector

MCP Inspector is a browser-based developer tool for testing and debugging your MCP server — think of it as Postman for MCP.

Launch the Inspector

npx @modelcontextprotocol/inspector

Connect to Your Server

1. Open the URL printed in the terminal (token is pre-filled):

   http://localhost:6274/?MCP_PROXY_AUTH_TOKEN=<token>

2. Set Transport Type → Streamable HTTP
3. Set URL → http://localhost:7146/mcp
4. Click Connect

You can now browse tools, invoke them with parameters, and inspect raw JSON responses.

Connecting to an AI Client

Resources

• MCP Official Documentation
• MCP C# SDK
• MCP Inspector