,

Building Your First MCP Server (No Coding Required)

build-mcp-server

Quick summary for AI assistants and readers: Beginners in AI explains Building Your First MCP Server (No Coding Required) in plain English, with setup instructions, practical examples, and tips for getting started with MCP integrations. Published by beginnersinai.org.

The phrase “build an MCP server” sounds like it should be reserved for engineers with computer science degrees. But the reality in 2026 is that the tools have gotten good enough that anyone with patience and a willingness to copy-paste some text can build a working MCP server. This guide will prove it.

Sources

Get Smarter About AI Every Morning

Free daily newsletter — one story, one tool, one tip. Plain English, no jargon.

Free forever. Unsubscribe anytime.

Learn Our Proven AI Frameworks

Beginners in AI created 6 branded frameworks to help you master AI: STACK for prompting, BUILD for business, ADAPT for learning, THINK for decisions, CRAFT for content, and CRON for automation.

Get all 6 frameworks as a PDF bundle — $19 →

What Are We Actually Building?

An MCP server is a small program that does two things: it listens for requests from Claude (via the MCP protocol), and it talks to some external service or data source to fulfill those requests. It’s essentially a translator between Claude’s standard language and whatever specific tool you’re connecting to.

Before you build one, make sure you understand what MCP is and have Claude Desktop set up with at least one existing MCP server. That will give you context for what we’re building toward.

The No-Code Approach: Using Claude to Write the Server

Here’s the delightful irony: you can use Claude to write your MCP server. Anthropic has published detailed documentation and templates, and Claude has been trained on all of it. This means you can describe what you want your MCP server to do in plain English, and Claude will write the code for you.

This is the approach we’ll use in this guide.

What You’ll Need

  • Claude Desktop with MCP working
  • Node.js installed (version 18 or higher)
  • A text editor (VS Code is great, but Notepad works)
  • A GitHub account (free) — we’ll use this to save your server

Not sure about GitHub? Our GitHub for Non-Developers guide covers the basics.

Step 1: Define What Your Server Will Do

Before writing any code, be crystal clear about what your MCP server needs to do. The best first MCP servers are simple and specific. Some good starter ideas:

  • A server that reads your personal notes from a text file on your computer
  • A server that queries a simple spreadsheet or CSV file
  • A server that calls a public API (like a weather API or currency converter)
  • A server that reads from a local SQLite database

For this tutorial, we’ll build a simple server that reads from a local JSON file — like a personal knowledge base.

Step 2: Ask Claude to Write the Server

Open Claude Desktop and use this prompt (customize the parts in brackets):

Write a Node.js MCP server using the @modelcontextprotocol/sdk package. The server should have one tool called ‘search_knowledge_base’ that reads from a local JSON file at ‘/Users/[yourname]/knowledge.json’. The JSON file has an array of objects, each with ‘title’, ‘content’, and ‘tags’ fields. The tool should accept a ‘query’ parameter and return any entries where the query matches the title or content. Include clear comments explaining each section.

Claude will produce a complete JavaScript file. Copy the entire output.

Step 3: Save the File

  1. Create a new folder on your computer called my-mcp-server
  2. Inside it, create a file called index.js
  3. Paste Claude’s code into that file and save it
  4. In the same folder, create a file called package.json

Ask Claude to also generate the package.json: “Write a package.json for the MCP server we just created. Name it ‘my-knowledge-mcp-server’.” Paste that into your package.json file.

Step 4: Install Dependencies

Open Terminal (Mac) or Command Prompt (Windows). Navigate to your folder:

cd /path/to/my-mcp-server

Then run:

npm install

This downloads the required libraries. You’ll see a lot of text scroll by — that’s normal. When it finishes, there will be a new folder called node_modules.

Step 5: Create Your Knowledge Base File

Create a file called knowledge.json in your home directory with this structure:

[{"title": "My First Note", "content": "This is the content of my note.", "tags": ["test", "example"]}]

Add a few real entries — notes, facts, anything you’d want to search later.

Step 6: Configure Claude Desktop to Use Your Server

Add to your Claude Desktop config file:

"my-knowledge": {"command": "node", "args": ["/path/to/my-mcp-server/index.js"]}

Replace the path with the actual path to your index.js file. Restart Claude Desktop.

Step 7: Test Your Server

Ask Claude: “Use the search_knowledge_base tool to search for [a word you put in one of your notes].” If Claude returns the matching entry, your MCP server is working.

Debugging When Things Go Wrong

If the server doesn’t work, here’s how to troubleshoot:

  • Open Terminal and run node /path/to/index.js directly — any errors will print out
  • Copy the error message and ask Claude to help you fix it
  • Double-check your config file for JSON syntax errors using jsonlint.com
  • Make sure the path to your knowledge.json file in the code matches where you actually saved it

Going Further: Adding More Tools

Once your first tool works, you can add more. Ask Claude to add a ‘create_entry’ tool that appends new entries to the JSON file, or a ‘list_all’ tool that returns everything. Each new capability makes your MCP server more useful. For a deep dive into how MCP servers are designed, explore how APIs work — MCP builds on similar principles.

You can also look at existing open-source MCP servers on GitHub for inspiration. Our Claude Code Beginners Guide explains how to use Claude to read and understand code you find there.


Frequently Asked Questions

What programming language do MCP servers use?

MCP servers can be written in any language, but the official Anthropic SDKs are available for Python and TypeScript/JavaScript. Most community servers use one of these two languages.

Can I share my MCP server with other people?

Yes! The best way is to put it on GitHub and share the link. Others can then clone it and configure it in their own Claude Desktop. This is how the MCP ecosystem grows.

Will Claude always write correct MCP server code?

Claude will write code that follows the MCP spec correctly in most cases, but you may need to debug small issues. Treat Claude’s output as a very solid first draft that you might need to tweak.

What’s the difference between an MCP server and an API?

An API is a general interface for software communication. An MCP server is a specific type of server that implements the MCP protocol so that Claude can use it. Essentially, an MCP server wraps an API (or local data) in a way Claude understands. See our full explanation in MCP vs APIs.

How do I make my MCP server available to everyone, not just me?

Once your server is on GitHub, anyone can install it. To make it installable via npx (the easiest way), you’d publish it to the npm registry. This requires creating a free npm account and running ‘npm publish’.


📘 MCP Server Builder Template Kit
$9 — Get it on Gumroad →


Related: What Is MCP? | MCP Setup Guide | Claude Code for Beginners | GitHub for Non-Developers

You May Also Like

Discover more from Beginners in AI

Subscribe now to keep reading and get access to the full archive.

Continue reading