MCP TypeScript Server Template
A template repository for building Model Context Protocol (MCP) servers with TypeScript, featuring full TypeScript support, testing setup, CI/CD pipelines, and modular architecture for easy extension.
README
MCP MCP Server
An MCP (Model Context Protocol) server implementation with TypeScript.
Features
- 🚀 Full TypeScript support with strict mode
- 🧪 Testing setup with Vitest and coverage reporting
- 📦 Automated releases with semantic-release
- 🔄 CI/CD pipelines with GitHub Actions
- 🏗️ Modular architecture for easy extension
- 📝 Comprehensive documentation and examples
- 🛠️ Development tools: Biome for linting/formatting, Husky for Git hooks
- 🎯 Pre-configured for MCP server development
- 🔐 Built-in validation using Zod schemas
- ⚡ ES modules with native Node.js support
- 📊 Code coverage reporting with c8
MCP Servers Built with This Template
Here are some MCP servers built using this template:
Wayback Machine MCP
Archive and retrieve web pages using the Internet Archive's Wayback Machine. No API keys required.
OpenAlex MCP
Access scholarly articles and research data from the OpenAlex database.
Building an MCP server? Use this template and add your server to this list!
Quick Start
Using GitHub Template
- Click "Use this template" button on GitHub
- Clone your new repository
- Install dependencies:
yarn install - Start development:
yarn dev
Manual Setup
# Clone the template
git clone https://github.com/Mearman/mcp-template.git my-mcp-server
cd my-mcp-server
# Install dependencies
yarn install
# Start development
yarn dev
Project Structure
src/
├── index.ts # MCP server entry point
├── tools/ # Tool implementations
│ ├── example.ts # Example tool
│ └── *.test.ts # Tool tests
├── utils/ # Shared utilities
│ ├── validation.ts # Input validation helpers
│ └── *.test.ts # Utility tests
└── types.ts # TypeScript type definitions
# Configuration files
├── .github/workflows/ # CI/CD pipelines
├── .husky/ # Git hooks
├── biome.json # Linter/formatter config
├── tsconfig.json # TypeScript config
├── vitest.config.ts # Test runner config
└── .releaserc.json # Semantic release config
Development
Available Commands
# Install dependencies
yarn install
# Development with hot reload
yarn dev
# Build TypeScript to JavaScript
yarn build
# Run production build
yarn start
# Run tests
yarn test
# Run tests in watch mode
yarn test:watch
# Run tests with coverage report
yarn test:ci
# Lint code
yarn lint
# Auto-fix linting issues
yarn lint:fix
# Format code
yarn format
Development Workflow
- Start development:
yarn dev- Runs the server with hot reload - Write tests: Add
.test.tsfiles alongside your code - Run tests:
yarn test:watch- Keep tests running while you code - Lint/format: Automatic on commit via Husky hooks
Creating Your MCP Server
1. Define Your Tools
Create tool implementations in src/tools/:
// src/tools/my-tool.ts
import { z } from 'zod';
const MyToolSchema = z.object({
input: z.string().describe('Tool input'),
});
export async function myTool(args: unknown) {
const { input } = MyToolSchema.parse(args);
// Tool implementation
return {
success: true,
result: `Processed: ${input}`,
};
}
2. Register Tools in Server
Update src/index.ts to register your tools:
server.setRequestHandler(ListToolsRequestSchema, async () => ({
tools: [
{
name: 'my_tool',
description: 'Description of what my tool does',
inputSchema: zodToJsonSchema(MyToolSchema),
},
],
}));
server.setRequestHandler(CallToolRequestSchema, async (request) => {
switch (request.params.name) {
case 'my_tool':
return await myTool(request.params.arguments);
default:
throw new Error(`Unknown tool: ${request.params.name}`);
}
});
3. Configure for Claude Desktop
Add to claude_desktop_config.json:
{
"mcpServers": {
"my-mcp-server": {
"command": "node",
"args": ["/path/to/my-mcp-server/dist/index.js"],
"env": {}
}
}
}
Testing
Write tests for your tools in src/tools/*.test.ts:
import { describe, it, expect } from 'vitest';
import { myTool } from './my-tool';
describe('myTool', () => {
it('should process input correctly', async () => {
const result = await myTool({ input: 'test' });
expect(result.success).toBe(true);
expect(result.result).toBe('Processed: test');
});
});
Publishing
NPM Package
-
Update
package.jsonwith your package details:name: Your package name (e.g.,mcp-your-server)description: Clear description of what your server doeskeywords: Add relevant keywords for discoverabilityauthor: Your name or organizationrepository: Your GitHub repository URL
-
Build the project:
yarn build -
Test the build locally:
yarn start -
Publish to npm:
npm publish
Automated Releases
This template includes semantic-release for automated versioning and publishing:
- Follow conventional commits
- Push to main branch
- CI/CD will automatically:
- Determine version bump
- Update CHANGELOG.md
- Create GitHub release
- Publish to npm (if NPM_TOKEN secret is configured)
Note: NPM publishing is optional. If you don't want to publish to npm, simply don't add the NPM_TOKEN secret to your repository. The release process will still create GitHub releases.
Best Practices
- Input Validation: Always validate tool inputs using Zod schemas
- Error Handling: Provide clear error messages for debugging
- Testing: Write comprehensive tests for all tools
- Documentation: Document each tool's purpose and usage
- Type Safety: Leverage TypeScript's type system fully
- Modular Design: Keep tools focused on single responsibilities
- Async/Await: Use modern async patterns for all I/O operations
- Environment Variables: Use
.envfiles for configuration (never commit secrets)
Adding CLI Support
To add CLI functionality to your MCP server (like the Wayback Machine example):
-
Install Commander.js:
yarn add commander chalk ora -
Create
src/cli.ts:import { Command } from 'commander'; import chalk from 'chalk'; export function createCLI() { const program = new Command(); program .name('your-tool') .description('Your MCP server as a CLI') .version('1.0.0'); // Add commands here return program; } -
Update
src/index.tsto detect CLI mode:async function main() { const isCliMode = process.stdin.isTTY || process.argv.length > 2; if (isCliMode && process.argv.length > 2) { const { createCLI } = await import('./cli.js'); const program = createCLI(); await program.parseAsync(process.argv); } else { // MCP server mode const transport = new StdioServerTransport(); await server.connect(transport); } } -
Add bin entry to
package.json:"bin": { "your-tool": "dist/index.js" }
License
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
Maintaining Your MCP Server
Template Updates
This template includes scripts to help you merge updates from the template into your derived MCP server:
-
Merge Updates: Use
scripts/merge-template-updates.shto selectively merge template improvements while preserving your customizations. -
Analyze Differences: Use
scripts/analyze-template-diff.shto understand what has changed between your server and the template.
See scripts/README.md for detailed documentation on using these maintenance scripts.
Installing Scripts in Existing Projects
If you created your MCP server before these scripts were added:
# From mcp-template directory
./scripts/install-scripts.sh ../path-to-your-mcp-server
Contributing
Contributions are welcome! Please read our contributing guidelines and submit pull requests to our repository.
Troubleshooting
Common Issues
- Build errors: Ensure all dependencies are installed with
yarn install - Type errors: Run
npx tsc --noEmitto check TypeScript types - Test failures: Check test files are named
*.test.ts - Claude Desktop connection: Verify the path in your config is absolute
Debug Mode
To see detailed logs when running as an MCP server:
DEBUG=* node dist/index.js
Resources
- Model Context Protocol Documentation
- MCP TypeScript SDK
- Creating MCP Servers Guide
- Awesome MCP Servers - Community-curated list
- MCP Discord - Get help and share your servers
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。