Open-source AI coworkers that each get a computer of their own: a browser, files and tools, with every action decided before it happens and recorded after. Bring any AG-UI agent.
MCP server template for FastMCP projects
This repo is a starter project for building your own MCP server with FastMCP. It gives you the basic server structure, transport options for stdio and HTTP, and TypeScript setup so you can begin adding MCP tools, resources, and prompts right away.
Builders who want a ready-made starting point for a custom MCP server.
You can start a new MCP server without assembling the project structure yourself.
What it does
FastMCP starter
Uses FastMCP as the base framework for the server.
Dual transport support
Includes stdio mode and HTTP/SSE mode for running the server locally or over the network.
TypeScript setup
Ships with TypeScript configuration and TypeScript source files.
Extensible MCP structure
Provides example shapes for tools, resources, and prompts.
Project creation commands
Creates a new server project from the template with `npx @mcpdotdirect/create-mcp-server` or `npm init @mcpdotdirect/mcp-server`.
How to get it
- 1Run
# with npx npx @mcpdotdirect/create-mcp-server # Or with npm npm init @mcpdotdirect/mcp-server
- 2Start the server
# Start the stdio server npm start # Or start the HTTP server npm run start:http
- 3For development with auto-reload
# Development mode with stdio npm run dev # Development mode with HTTP npm run dev:http
- 4By default, the HTTP server runs on port 3001. You can change this by setting the PORT…
# Start the HTTP server on a custom port PORT=8080 npm run start:http
- 5FastMCP provides built-in tools for testing your server
# Test with mcp-cli npx fastmcp dev server.js # Inspect with MCP Inspector npx fastmcp inspect server.ts
- 6You can customize the server using environment variables
# Change the HTTP port (default is 3001) PORT=8080 npm run start:http # Change the host binding (default is 0.0.0.0) HOST=127.0.0.1 npm run start:http
README
@mcpdotdirect/template-mcp-server
A CLI tool to quickly get started building your very own MCP (Model Context Protocol) server using FastMCP
📋 Usage
# with npx
npx @mcpdotdirect/create-mcp-server
# Or with npm
npm init @mcpdotdirect/mcp-server
🔭 What's Included
The template includes:
- Basic server setup with both stdio and HTTP transport options using FastMCP
- Structure for defining MCP tools, resources, and prompts
- TypeScript configuration
- Development scripts and configuration
✨ Features
- FastMCP: Built using the FastMCP framework for simpler implementation
- Dual Transport Support: Run your MCP server over stdio or HTTP
- TypeScript: Full TypeScript support for type safety
- Extensible: Easy to add custom tools, resources, and prompts
🚀 Getting Started
After creating your project:
-
Install dependencies using your preferred package manager:
# Using npm npm install # Using yarn yarn # Using pnpm pnpm install # Using bun bun install -
Start the server:
# Start the stdio server npm start # Or start the HTTP server npm run start:http -
For development with auto-reload:
# Development mode with stdio npm run dev # Development mode with HTTP npm run dev:http
Note: The default scripts in package.json use Bun as the runtime (e.g.,
bun run src/index.ts). If you prefer to use a different package manager or runtime, you can modify these scripts in your package.json file to use Node.js or another runtime of your choice.
📖 Detailed Usage
Transport Methods
The MCP server supports two transport methods:
-
stdio Transport (Command Line Mode):
- Runs on your local machine
- Managed automatically by Cursor
- Communicates directly via
stdout - Only accessible by you locally
- Ideal for personal development and tools
-
SSE Transport (HTTP Web Mode):
- Can run locally or remotely
- Managed and run by you
- Communicates over the network
- Can be shared across machines
- Ideal for team collaboration and shared tools
Running the Server Locally
stdio Transport (CLI Mode)
Start the server in stdio mode for CLI tools:
# Start the stdio server
npm start
# or with other package managers
yarn start
pnpm start
bun start
# Start the server in development mode with auto-reload
npm run dev
# or
yarn dev
pnpm dev
bun dev
HTTP Transport (Web Mode)
Start the server in HTTP mode for web applications:
# Start the HTTP server
npm run start:http
# or
yarn start:http
pnpm start:http
bun start:http
# Start the HTTP server in development mode with auto-reload
npm run dev:http
# or
yarn dev:http
pnpm dev:http
bun dev:http
By default, the HTTP server runs on port 3001. You can change this by setting the PORT environment variable:
# Start the HTTP server on a custom port
PORT=8080 npm run start:http
Connecting to the Server
Connecting from Cursor
To connect to your MCP server from Cursor:
- Open Cursor and go to Settings (gear icon in the bottom left)
- Click on "Features" in the left sidebar
- Scroll down to "MCP Servers" section
- Click "Add new MCP server"
- Enter the following details:
- Server name:
my-mcp-server(or any name you prefer) - For stdio mode:
- Type:
command - Command: The path to your server executable, e.g.,
npm start
- Type:
- For SSE mode:
- Type:
url - URL:
http://localhost:3001/sse
- Type:
- Server name:
- Click "Save"
Using mcp.json with Cursor
For a more portable configuration, create an .cursor/mcp.json file in your project's root directory:
{
"mcpServers": {
"my-mcp-stdio": {
"command": "npm",
"args": [
"start"
],
"env": {
"NODE_ENV": "development"
}
},
"my-mcp-sse": {
"url": "http://localhost:3001/sse"
}
}
}
You can also create a global configuration at ~/.cursor/mcp.json to make your MCP servers available in all your Cursor workspaces.
Note:
- The
commandtype entries run the server in stdio mode - The
urltype entry connects to the HTTP server using SSE transport - You can provide environment variables using the
envfield - When connecting via SSE with FastMCP, use the full URL including the
/ssepath:http://localhost:3001/sse
Testing Your Server with CLI Tools
FastMCP provides built-in tools for testing your server:
# Test with mcp-cli
npx fastmcp dev server.js
# Inspect with MCP Inspector
npx fastmcp inspect server.ts
Using Environment Variables
You can customize the server using environment variables:
# Change the HTTP port (default is 3001)
PORT=8080 npm run start:http
# Change the host binding (default is 0.0.0.0)
HOST=127.0.0.1 npm run start:http
🛠️ Adding Custom Tools and Resources
When adding custom tools, resources, or prompts to your FastMCP server:
Tools
server.addTool({
name: "hello_world",
description: "A simple hello world tool",
parameters: z.object({
name: z.string().describe("Name to greet")
}),
execute: async (params) => {
return `Hello, ${params.name}!`;
}
});
Resources
server.addResourceTemplate({
uriTemplate: "example://{id}",
name: "Example Resource",
mimeType: "text/plain",
arguments: [
{
name: "id",
description: "Resource ID",
required: true,
},
],
async load({ id }) {
return {
text: `This is an example resource with ID: ${id}`
};
}
});
Prompts
server.addPrompt({
name: "greeting",
description: "A simple greeting prompt",
arguments: [
{
name: "name",
description: "Name to greet",
required: true,
},
],
load: async ({ name }) => {
return `Hello, ${name}! How can I help you today?`;
}
});
📚 Documentation
For more information about FastMCP, visit FastMCP GitHub Repository.
For more information about the Model Context Protocol, visit the MCP Documentation.
📄 License
This project is licensed under the MIT License - see the LICENSE file for details.
Files in the repo
- .github
- bin
- src
- .gitignore
- CHANGELOG.md
- LICENSE
- package.json
- README.md
- tsconfig.json
Discussion (0)
Ask about usage, or say what you built with itSign in to join the discussion.
No comments yet. Be the first to say what this is good for.
More templates
Turn Claude Code into a full game dev studio — 49 AI agents, 72 workflow skills, and a complete coordination system mirroring real studio hierarchy.
A self-organizing Obsidian vault that gives AI coding agents persistent memory. Claude Code, Codex CLI, Gemini CLI.
Production-grade Playwright + TypeScript Scaffold for Agentic Testing. Harness for all major AI coding agents baked in.
Omnia Vault - the all-in-one project brain: an Obsidian LLM wiki, Graphify code graphs, a living plan that triages new videos against itself, and a Claude Code ⇄ Codex relay. Everything your project knows, in one clonable vault.

🎬 Tạo video "so sánh kiến thức" ngắn tự động — HyperFrames + AI voice, 1 template nhiều chủ đề.