template-mcp
A TypeScript MCP server template with Zod validation, dual transport (stdio/HTTP), and modular architecture for building MCP-compatible tools, resources, and prompts.
README
template-mcp
MCP (Model Context Protocol) server template with TypeScript, Zod validation, and dual transport support (stdio/HTTP). Compatible with any MCP client: Claude Code, Claude Desktop, Cursor, VS Code Copilot, Windsurf, Cline, and more.
Features
- Dual transport: stdio (local) and Streamable HTTP (remote)
- TypeScript strict with ESM modules
- Zod validation for tool input schemas
- Joi env validation (fail-fast on startup)
- Pino logging to stderr (stdio-safe)
- Modular architecture: tools, resources, and prompts as separate modules
- Factory pattern:
createServer()for testability - Full test suite with MCP SDK in-memory transport
- Quality tooling: ESLint + Prettier + Husky + lint-staged
- Docker ready: multi-stage build
- CI/CD: GitHub Actions pipeline
Quick Start
pnpm install
pnpm dev
Scripts
| Script | Description |
|---|---|
pnpm dev |
Start with hot reload (tsx watch) |
pnpm build |
Compile TypeScript + resolve aliases |
pnpm start |
Run compiled server |
pnpm test |
Run tests |
pnpm lint |
Lint source code |
pnpm type-check |
Type check without emit |
Configuration
Copy .env.example to .env and adjust:
| Variable | Default | Description |
|---|---|---|
MCP_TRANSPORT |
stdio |
Transport: stdio or http |
PORT |
3000 |
HTTP port (only for http transport) |
LOG_LEVEL |
info |
Pino log level |
NODE_ENV |
development |
Environment |
Project Structure
src/
├── main.ts # Entrypoint: transport selection
├── server.ts # createServer() factory
├── config/ # Env validation + constants
├── common/ # Logger, error helpers, types
├── tools/ # MCP tools (callable by LLMs)
├── resources/ # MCP resources (read-only data)
└── prompts/ # MCP prompts (reusable templates)
Adding a New Tool
- Create
src/tools/my-tool.tool.ts:
import type { McpServer } from '@modelcontextprotocol/sdk/server/mcp.js';
import { z } from 'zod';
export function registerMyTool(server: McpServer): void {
server.registerTool(
'my_tool',
{
title: 'My Tool',
description: 'What this tool does',
inputSchema: {
param: z.string().describe('Parameter description'),
},
annotations: {
readOnlyHint: true,
destructiveHint: false,
idempotentHint: true,
openWorldHint: false,
},
},
async ({ param }) => ({
content: [{ type: 'text', text: `Result: ${param}` }],
}),
);
}
- Register in
src/tools/index.ts:
import { registerMyTool } from './my-tool.tool.js';
export function registerTools(server: McpServer): void {
registerGreetTool(server);
registerMyTool(server); // add here
}
- Add tests in
src/tools/__tests__/my-tool.tool.spec.ts
Client Configuration
Claude Code
Add to .claude/settings.json:
{
"mcpServers": {
"template-mcp": {
"command": "node",
"args": ["/absolute/path/to/template-mcp/dist/main.js"]
}
}
}
Claude Desktop
Add to claude_desktop_config.json:
{
"mcpServers": {
"template-mcp": {
"command": "node",
"args": ["/absolute/path/to/template-mcp/dist/main.js"]
}
}
}
Cursor
Add to Cursor Settings > MCP Servers:
{
"mcpServers": {
"template-mcp": {
"command": "node",
"args": ["/absolute/path/to/template-mcp/dist/main.js"]
}
}
}
VS Code (Copilot)
Add to .vscode/settings.json:
{
"mcp": {
"servers": {
"template-mcp": {
"command": "node",
"args": ["/absolute/path/to/template-mcp/dist/main.js"]
}
}
}
}
Docker
# Build
docker build -t template-mcp .
# Run (HTTP mode, used for remote access)
docker run -p 3000:3000 template-mcp
Tech Stack
- Node.js 22 + TypeScript (strict, ESM)
- MCP SDK v1 (
@modelcontextprotocol/sdk) - Zod (tool input validation)
- Joi (env validation)
- Pino (stderr logging)
- Vitest (testing)
- ESLint + Prettier + Husky
Verificación
Todo lo siguiente está comprobado y funcionando al 100%.
Calidad de código
| Check | Comando |
|---|---|
| Lint + formato | pnpm lint |
| Tipado estricto | pnpm type-check |
| Build (tsc + alias) | pnpm build |
Tests unitarios (11/11)
pnpm test
| Suite | Cubre |
|---|---|
greet.tool.spec.ts (5) |
Listado, estilos casual/formal/enthusiastic, rechazo de nombre vacío |
server-info.resource.spec.ts (2) |
Listado, campos JSON (name, version, uptime, timestamp) |
summarize.prompt.spec.ts (4) |
Listado, estilos brief/bullet-points, coerción numérica, defaults |
Sin red ni puertos — usa InMemoryTransport del SDK.
Runtime — transporte stdio (modo por defecto)
echo '{"jsonrpc":"2.0","id":1,"method":"initialize","params":{"protocolVersion":"2024-11-05","capabilities":{},"clientInfo":{"name":"test","version":"1.0.0"}}}' \
| MCP_TRANSPORT=stdio node dist/main.js
Respuesta JSON-RPC en stdout, logs en stderr.
Runtime — transporte HTTP
MCP_TRANSPORT=http PORT=3100 node dist/main.js &
# Initialize → capturar Mcp-Session-Id del header
# tools/list, resources/list, prompts/list, tools/call greet, resources/read info://server
| Endpoint verificado | Resultado esperado |
|---|---|
tools/call greet {"name":"Freddy","style":"casual"} |
"Hey Freddy! How's it going?" |
resources/read info://server |
JSON con name, version, uptime, nodeVersion, timestamp |
Docker
docker build -t template-mcp . # multi-stage: base → deps → build → production
docker run -p 3000:3000 template-mcp # arranca en HTTP mode
CI (GitHub Actions)
pnpm install → pnpm lint → pnpm build → pnpm test
Corre en cada push y PR a main/master.
Pipeline de commits (local)
git commit → husky → lint-staged → eslint --fix + prettier --write (solo archivos staged)
Gaps conocidos
- HTTP transport sin tests automáticos (medio): los unit tests usan
InMemoryTransport; el transporte HTTP (StreamableHTTPServerTransport) solo se verificó manualmente con curl. Para producción remota añadir tests de integración con sesión real - Integración con cliente MCP (medio): verificar manualmente agregando a
.claude/settings.jsono Cursor y confirmando que tools/resources/prompts aparecen en el cliente - Tool inputs extremos (bajo): strings muy largos, unicode malformado — Zod los rechaza pero el error response no está testeado vía HTTP
- Sesiones concurrentes (bajo): fuera del scope de un template
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.