@webbywisp/create-mcp-server

Scaffold production-ready MCP (Model Context Protocol) servers in seconds.

npx @webbywisp/create-mcp-server my-server

Built for developers who want to expose tools, resources, or prompt templates to AI models — without writing boilerplate.

Quick Start

# Interactive wizard (recommended)
npx @webbywisp/create-mcp-server my-server

# Non-interactive, use defaults (tool-server template)
npx @webbywisp/create-mcp-server my-server --yes

# Pick a specific template
npx @webbywisp/create-mcp-server my-server -t resource-server

# Preview what would be created
npx @webbywisp/create-mcp-server my-server --dry-run

Commands

create-mcp-server [directory]           Scaffold a new MCP server (default)
create-mcp-server init [directory]      Scaffold a new MCP server
create-mcp-server add [type] [name]     Add a tool/resource/prompt to an existing server
create-mcp-server list                  Show available templates

Templates

tool-server (default)

Expose callable tools to AI models. Great for file I/O, HTTP requests, shell commands, APIs.

Generated tools:

• read_file — read file contents with encoding + size limits
• write_file — write files, creating directories as needed
• http_fetch — make HTTP requests (GET, POST, PUT, DELETE)
• exec_command — execute shell commands with safety checks

npx @webbywisp/create-mcp-server my-tools -t tool-server

resource-server

Expose data as addressable URIs. Great for files, databases, config, any content an AI should be able to read.

Generated resources:

• file:// — list and read files from the filesystem
• db:// — query a SQLite database
• config:// — expose configuration files

npx @webbywisp/create-mcp-server my-data -t resource-server

prompt-server

Expose structured prompt templates for common tasks. Great for standardizing code review, debugging, documentation, refactoring workflows.

Generated prompts:

• code-review — review code for issues, patterns, suggestions
• debug-error — systematic error analysis
• generate-docs — generate documentation from code
• refactor — refactor code with specific goals

npx @webbywisp/create-mcp-server my-prompts -t prompt-server

Options

-t, --template <name>    Template: tool-server | resource-server | prompt-server
-y, --yes                Skip prompts, use defaults
--dry-run                Preview what would be created without writing files
--force                  Overwrite existing files

Generated Structure

tool-server

my-server/
├── src/
│   ├── index.ts              # Server entry point, tool registration
│   └── tools/
│       ├── readFile.ts       # Read file contents
│       ├── writeFile.ts      # Write files
│       ├── httpFetch.ts      # HTTP requests
│       └── execCommand.ts    # Shell execution (with safety checks)
├── package.json              # MCP SDK + zod dependencies
├── tsconfig.json             # TypeScript config (ES2022, NodeNext)
├── .gitignore
└── README.md

resource-server

my-server/
├── src/
│   ├── index.ts              # Server entry point, resource registration
│   └── resources/
│       ├── fileSystem.ts     # file:// URI resources
│       ├── database.ts       # db:// URI resources (SQLite)
│       └── config.ts         # config:// URI resources
├── package.json
├── tsconfig.json
├── .gitignore
└── README.md

prompt-server

my-server/
├── src/
│   ├── index.ts              # Server entry point, prompt registration
│   └── prompts/
│       ├── codeReview.ts     # Code review prompt
│       ├── debugError.ts     # Error debugging prompt
│       ├── generateDocs.ts   # Documentation generation prompt
│       └── refactor.ts       # Refactoring prompt
├── package.json
├── tsconfig.json
├── .gitignore
└── README.md

After Scaffolding

cd my-server
npm install
npm run build
npm start

Connect to Claude Desktop

Edit ~/Library/Application Support/Claude/claude_desktop_config.json:

{
  "mcpServers": {
    "my-server": {
      "command": "node",
      "args": ["/absolute/path/to/my-server/dist/index.js"]
    }
  }
}

Restart Claude Desktop. Your tools/resources/prompts will be available immediately.

Connect to Other Clients

The stdio transport works with any MCP-compliant client. See the MCP docs for SSE transport and other options.

Adding to an Existing Server

The add command lets you add individual pieces to a server you've already scaffolded:

# Add a new tool file
create-mcp-server add tool my-custom-tool

# Add a new resource handler  
create-mcp-server add resource my-data-source

# Add a new prompt template
create-mcp-server add prompt code-explainer

Why MCP?

The Model Context Protocol (MCP) is an open standard for connecting AI models to external tools and data. Instead of hard-coding API calls inside prompts, MCP lets you:

• Expose tools — functions AI can call (file ops, HTTP, databases, APIs)
• Expose resources — data AI can read on demand (files, DB records, config)
• Expose prompts — reusable prompt templates with structured inputs

AI models discover what's available and call your code directly. Clean separation, standardized interface, works across Claude, OpenAI, and any MCP-compatible client.

This CLI scaffolds the full TypeScript setup so you can focus on the logic, not the boilerplate.

Related

• @webbywisp/create-ai-agent — scaffold AI agent workspaces (SOUL.md, OPS.md, memory structure)
• MCP Specification
• AI Agent Workspace Kit — full production agent workspace template ($19)

License

MIT