{"id":1254,"date":"2026-04-06T10:53:27","date_gmt":"2026-04-06T17:53:27","guid":{"rendered":"https:\/\/www.kenwalger.com\/blog\/?p=1254"},"modified":"2026-03-30T08:28:01","modified_gmt":"2026-03-30T15:28:01","slug":"mcp-quickstart-typescript-vs-python","status":"publish","type":"post","link":"https:\/\/www.kenwalger.com\/blog\/ai\/mcp-quickstart-typescript-vs-python\/","title":{"rendered":"Building Your First MCP Server: TypeScript vs. Python"},"content":{"rendered":"

The 5-Minute “Hello World” Comparison<\/h3>\n

We\u2019ve spent the last month talking about the End of Glue Code<\/a> and the Enterprise AI Mesh<\/a>. But if you\u2019re a developer, you don’t just want to see the blueprint\u2014you want to hold the tools.<\/p>\n

Whether you are a TypeScript<\/a> veteran or a Python<\/a> enthusiast, building an MCP server is surprisingly simple. Today, we\u2019re going to build the same “Hello World” tool in both languages to show you exactly how the protocol abstracts away the complexity.<\/p>\n

1. The TypeScript Approach (Node.js)<\/h2>\n

TypeScript is the “native” language of the Model Context Protocol, and the @modelcontextprotocol\/sdk<\/code> is exceptionally robust for high-performance enterprise tools.<\/p>\n

Prerequisites:<\/strong><\/p>\n

npm install @modelcontextprotocol\/sdk zod\n<\/code><\/pre>\n
The Code:<\/strong><\/p>\n
import { Server } from \"@modelcontextprotocol\/sdk\/server\/index.js\";\nimport { StdioServerTransport } from \"@modelcontextprotocol\/sdk\/server\/stdio.js\";\nimport { z } from \"zod\";\n\nconst server = new Server({\n  name: \"hello-world-server\",\n  version: \"1.0.0\",\n}, {\n  capabilities: { tools: {} }\n});\n\n\/\/ Define a simple greeting tool\nserver.tool(\n  \"greet_user\",\n  { name: z.string().describe(\"The name of the person to greet\") },\n  async ({ name }) => {\n    return {\n      content: [{ type: \"text\", text: `Hello, ${name}! Welcome to the MCP Mesh.` }]\n    };\n  }\n);\n\nasync function main() {\n  const transport = new StdioServerTransport();\n  await server.connect(transport);\n}\n\nmain().catch(console.error);\n<\/code><\/pre>\n
2. The Python Approach<\/h2>\n
For data scientists and AI engineers, the Python SDK offers a beautifully decorative approach. It feels more “agent-native” and integrates seamlessly with existing AI libraries.<\/p>\n
Prerequisites:<\/strong><\/p>\n
pip install mcp\n<\/code><\/pre>\n
The Code:<\/strong><\/p>\n
import asyncio\nfrom mcp.server.fastmcp import FastMCP\n\n# Initialize FastMCP - the \"Quick Start\" wrapper\nmcp = FastMCP(\"HelloWorld\")\n\n@mcp.tool()\nasync def greet_user(name: str) -> str:\n    \"\"\"Greets a user by name.\"\"\"\n    return f\"Hello, {name}! Welcome to the MCP Mesh.\"\n\nif __name__ == \"__main__\":\n    mcp.run(transport='stdio')\n<\/code><\/pre>\n
Side-by-Side: Which Should You Choose?<\/h2>\n\n\n\n\n\n\n\n\n
Feature<\/th>\nTypeScript (Standard SDK)<\/th>\nPython (FastMCP)<\/th>\n<\/tr>\n<\/thead>\n
Best For<\/strong><\/td>\nHigh-performance, Type-safe tools<\/td>\nRapid prototyping, AI logic<\/td>\n<\/tr>\n
Validation<\/strong><\/td>\nZod (Explicit & Strict)<\/td>\nPydantic \/ Type Hints (Implicit)<\/td>\n<\/tr>\n
Verbosity<\/strong><\/td>\nModerate (Structured)<\/td>\nMinimal (Decorator-based)<\/td>\n<\/tr>\n
Transport<\/strong><\/td>\nSTDIO, SSE, Custom<\/td>\nSTDIO, SSE<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n
How to Test Your Server<\/h2>\n
Once you’ve saved your code, you don’t need a complex frontend to test it. Use the MCP Inspector:<\/strong><\/p>\n
# For TypeScript\nnpx @modelcontextprotocol\/inspector node build\/index.js\n\n# For Python\nnpx @modelcontextprotocol\/inspector python your_script.py\n<\/code><\/pre>\n
This will launch a local web interface where you can perform the “Protocol Handshake” and trigger your tools manually. It’s the best way to verify your “Zero-Glue” infrastructure before connecting it to an agent.<\/p>\n
Conclusion<\/h2>\n
The “Zero-Glue” architecture isn’t about which language you use\u2014it’s about the Protocol<\/em>. As you can see, the logic for the “Hello World” tool is nearly identical in both versions. The Model Context Protocol ensures that no matter how you build your tools, your agents can discover and use them in a standardized way.<\/p>\n
Ready to build your own?<\/h3>\n
Check out the reference repo for more complex examples, including Notion and Oracle 26ai integrations.<\/p>\n
MCP Forensic Analyzer Repository<\/a><\/p>\n
The “Zero-Glue” Series<\/h2>\n