One-Click Deploy an SSE MCP Server on Alibaba Cloud with Serverless Devs CLI

What is Model Context Protocol (MCP)?

Model Context Protocol (MCP) is a standardized protocol that connects large language model (LLM) capabilities with the real world, defining Resources, Prompts, and Tools to enable modular extensions such as file system access, API integration, and IoT control for LLMs.

Current pain points of MCP Server development

Developers must manually configure Python/TypeScript SDKs, resolve dependency conflicts, and handle complex IaC deployment on the cloud, making a zero‑to‑one MCP Server setup costly and fragmented.

Accelerating development with Serverless Devs CLI

Serverless Devs CLI provides one‑click project initialization and deployment, dramatically shortening the development workflow.

Initialize a MCP Server project

s init start-fcai-mcp-nodejs

Follow the prompts to set region, function name, and access configuration, which creates a ready‑to‑run Hello World MCP Server project.

Project directory structure

.
├── build.yaml          # Build configuration
├── code               # Project code
│   ├── check_node_version.mjs  # Pre‑deployment check script
│   ├── package-lock.json
│   ├── package.json
│   ├── src
│   │   └── index.ts   # Server source code
│   ├── tsconfig.json
│   └── webpack.config.js
├── readme.md
└── s.yaml             # Deployment YAML

Develop the MCP Server

import { McpServer } from "@modelcontextprotocol/sdk/server/mcp.js";
import { z } from "zod";
import { StdioServerTransport } from "@modelcontextprotocol/sdk/server/stdio.js";

const createServer = () => {
  const server = new McpServer({
    name: "my-mcp-server",
    version: "1.0.0",
  });

  server.tool(
    "hello_world",
    "Return string 'hello world!'",
    {},
    async () => {
      console.log("Hello World tool called");
      return { content: [{ type: "text", text: "hello world!" }] };
    }
  );
  return server;
};

const server = createServer();
const transport = new StdioServerTransport();
server.connect(transport).then(() => console.log("Server started")).catch(err => console.error("Error starting server:", err));

This server defines a simple hello_world tool that returns the string "hello world!" without any input parameters.

Package and deploy the server

npm install   # install dependencies if not done
npm run build  # bundle with webpack
s deploy -y   # one‑click deployment to Function AI

The deployment YAML ( s.yaml) generated by the CLI contains all required instance specifications, code paths, and trigger configurations.

Test the deployed MCP Server

Use the official inspector client (or the built‑in LLM‑enabled inspector) to connect to the server:

npx @modelcontextprotocol/inspector node build/index.js
npx @serverless-devs/inspector node build/index.js

Configure the internetUrl obtained after deployment, append /sse, and use the inspector UI to list tools, connect, and invoke the hello_world tool.

Client integration methods

Three ways to connect clients:

Official client (SSE) : Add the server URL with /sse to the client configuration JSON.

Custom local client : Use the SDK to create a client and an SSEClientTransport pointing to the deployed URL.

Remote client on Function Compute : Deploy the inspector as a function and access it via the provided domain.

Additional resources

Serverless Devs also offers ready‑made MCP Server templates for Python, Java, and many open‑source services (e.g., maps, GitHub). See the code repository for more examples.