MCP Server 6zn

🚀 What is MCP (Model Context Protocol)?

MCP (Model Context Protocol) is a standardized protocol that defines how AI models interact with applications. In simple terms, it's a specification that dictates how large language models (LLMs) and other AI models exchange data with clients (applications, services).

Key Concepts of MCP

MCP provides a communication structure for the interaction between AI models and data, similar to the concepts of the OpenAI API, LangChain, and function calling.

1. MCP Server (McpServer)

   • Create a server that can interact with AI models.
   • Define specific functions (tools, APIs) to handle client requests.

   import { McpServer } from "@modelcontextprotocol/sdk/server/mcp.js";
   
   const server = new McpServer({
     name: "My MCP Server",
     version: "1.0.0",
   });
   

2. MCP Tools (Tools)

   • Functions that AI models can call, similar to the concept of function calling.

   import { z } from "zod";
   
   server.tool(
     "add",
     {
       a: z.number(),
       b: z.number(),
     },
     async ({ a, b }) => {
       return {
         content: [{ type: "text", text: String(a + b) }],
       };
     }
   );
   

   • When the client calls the add tool, it returns the result of a + b.

3. MCP Resources (Resources)

   • Define specific data and enable clients to access this data.

   import { ResourceTemplate } from "@modelcontextprotocol/sdk/server/mcp.js";
   
   server.resource(
     "greeting",
     new ResourceTemplate("greeting://{name}", { list: undefined }),
     async (uri, { name }) => ({
       contents: [
         {
           uri: uri.href,
           text: `Hello, ${name}!`,
         },
       ],
     })
   );
   

   • When calling a URI like greeting://John, it returns "Hello, John!".

4. SSE (Server-Sent Events) Support

   • MCP also supports the SSE protocol for real-time data transfer.
   • It's very useful for handling streaming responses from large language models.

   import { SSEServerTransport } from "@modelcontextprotocol/sdk/server/sse.js";
   
   app.get("/sse", async (req, res) => {
     const transport = new SSEServerTransport("/messages", res);
     await server.connect(transport);
   });
   

   • When a client connects to the /sse endpoint, it can receive data in real-time.

💻 Usage Examples

Basic Usage

// Create an MCP server
import { McpServer } from "@modelcontextprotocol/sdk/server/mcp.js";

const server = new McpServer({
  name: "My MCP Server",
  version: "1.0.0",
});

// Define an MCP tool
import { z } from "zod";

server.tool(
  "add",
  {
    a: z.number(),
    b: z.number(),
  },
  async ({ a, b }) => {
    return {
      content: [{ type: "text", text: String(a + b) }],
    };
  }
);

// Define an MCP resource
import { ResourceTemplate } from "@modelcontextprotocol/sdk/server/mcp.js";

server.resource(
  "greeting",
  new ResourceTemplate("greeting://{name}", { list: undefined }),
  async (uri, { name }) => ({
    contents: [
      {
        uri: uri.href,
        text: `Hello, ${name}!`,
      },
    ],
  })
);

// Enable SSE support
import { SSEServerTransport } from "@modelcontextprotocol/sdk/server/sse.js";

app.get("/sse", async (req, res) => {
  const transport = new SSEServerTransport("/messages", res);
  await server.connect(transport);
});

📦 Where Can MCP Be Used?

• As an API server for LLMs and other AI models
• Building microservices that provide AI capabilities
• Supporting function calls for AI models
• As an alternative or supplementary solution for AI frameworks like LangChain
• Developing server - and client - based AI - driven conversational applications

📚 Conclusion

MCP (Model Context Protocol) is a standardized protocol that defines how AI models interact with applications. It allows us to easily implement features such as tools, resources, and streaming.

In a nutshell: 👉 MCP = A framework to help build AI model API servers! 🚀