my-mcp-server

my-mcp-server

A minimal starter template for building Model Context Protocol (MCP) servers using TypeScript and FastMCP, including an example weather tool for demonstration.

Category
访问服务器

README

<div align="center">

<img src="https://files.catbox.moe/vumztw.png" alt="ADK TypeScript Logo" width="100" />

<br/>

MCP Server Starter Template

A minimal starter template for building Model Context Protocol (MCP) servers using TypeScript and FastMCP.

MCP • FastMCP • TypeScript


</div>

A minimal starter template for building Model Context Protocol (MCP) servers using TypeScript and FastMCP.

Features

  • Basic project structure with src/lib, src/services, src/tools.
  • TypeScript setup (compiles to dist/).
  • fastmcp for MCP server implementation.
  • A weather service example demonstrating:
    • Proper folder structure (lib, services, tools)
    • API integration with error handling
    • Parameter validation using Zod
    • Separation of concerns
  • GitHub Actions workflows for CI and Release (manual trigger by default).

Getting Started

The easiest way to create a new MCP server project using this template is with the ADK CLI:

npm install -g @iqai/adk-cli # if you haven't already
adk new --template mcp-starter my-mcp-server
cd my-mcp-server
pnpm install

You can also use this template directly by copying the files, but using the CLI is recommended for best results.

Running the Server

Default (Production/Development) Route

To run your MCP server in production or for standard development, use:

pnpm dev

Fast Iteration & Agent Setup (ADK CLI)

For rapid prototyping, interactive testing, or initial agent setup, use the ADK CLI:

adk run   # Interactive CLI chat with your agents
adk web   # Web interface for easy testing and demonstration
  1. Configure environment variables: For the weather service example, you'll need an OpenWeather API key:

    # Create a .env file (add to .gitignore)
    echo "OPENWEATHER_API_KEY=your_api_key_here" > .env
    

    Get an API key from OpenWeather.

  2. Initial Commit: It's a good idea to make an initial commit at this stage.

    git add .
    git commit -m "feat: initial project setup from template"
    
  3. Develop your server:

    • Add your custom tools in the src/tools/ directory.
    • Implement logic in src/lib/ and src/services/.
    • Register tools in src/index.ts.

Example Weather Tool

This template includes a weather service example that demonstrates:

  1. HTTP Utilities (src/lib/http.ts):

    • Type-safe HTTP requests with Zod validation
    • Error handling
  2. Configuration (src/lib/config.ts):

    • Environment variable management
    • Service configuration
  3. Weather Service (src/services/weather-service.ts):

    • API integration
    • Data transformation
    • Proper error propagation
  4. Weather Tool (src/tools/weather.ts):

    • Parameter validation with Zod
    • User-friendly output formatting
    • Error handling and user guidance

To use the weather tool:

# Set your OpenWeather API key
export OPENWEATHER_API_KEY=your_api_key_here

# Run the server
pnpm run start

# Connect with an MCP client and use the GET_WEATHER tool
# with parameter: { "city": "London" }

Release Management (Changesets)

This template is ready for release management using Changesets.

  1. Install Changesets CLI (if not already in devDependencies): The template package.json should include @changesets/cli. If not:

    pnpm add -D @changesets/cli
    
  2. Initialize Changesets: This command will create a .changeset directory with some configuration files.

    pnpm changeset init
    # or npx changeset init
    

    Commit the generated .changeset directory and its contents.

  3. Adding Changesets During Development: When you make a change that should result in a version bump (fix, feature, breaking change):

    pnpm changeset add
    # or npx changeset add
    

    Follow the prompts. This will create a markdown file in the .changeset directory describing the change. Commit this changeset file along with your code changes.

  4. Publishing a Release: The GitHub Actions workflow release.yml (in mcp-server-starter/.github/workflows/) is set up for this. When you are ready to release:

    • Ensure all feature PRs with their changeset files are merged to main.
    • Important: Before publishing, ensure your package.json is complete. Add or update fields like keywords, author, repository (e.g., "repository": {"type": "git", "url": "https://github.com/YOUR_USERNAME/YOUR_REPO_NAME.git"}), bugs (e.g., "bugs": {"url": "https://github.com/YOUR_USERNAME/YOUR_REPO_NAME/issues"}), and homepage (e.g., "homepage": "https://github.com/YOUR_USERNAME/YOUR_REPO_NAME#readme") for better discoverability and information on npm.
    • The release.yml workflow (manually triggered by default in the template) will:
      1. Run changeset version to consume changeset files, update package.json versions, and update CHANGELOG.md. It will push these to a changeset-release/main branch and open a "Version Packages" PR.
      2. Merge the "Version Packages" PR.
      3. Upon merging, the workflow runs again on main. This time, it will run pnpm run publish-packages (which should include changeset publish) to publish to npm and create GitHub Releases/tags.
    • To enable automatic release flow: Change on: workflow_dispatch in release.yml to on: push: branches: [main] (or your release branch).

Available Scripts

  • pnpm run build: Compiles TypeScript to JavaScript in dist/ and makes the output executable.
  • pnpm run dev: Runs the server in development mode using tsx (hot-reloading for TypeScript).
  • pnpm run start: Runs the built server (from dist/) using Node.

Rapid Development Setup

For fast iteration during development, use the dev script with hot reload:

pnpm run dev

This runs tsx watch src/index.ts which automatically restarts the server when you make changes to your TypeScript files.

MCP Client Configuration for Development

Cursor (.cursor/mcp.json):

{
  "mcpServers": {
    "weather-dev": {
      "command": "/Users/marvelsmbp/Documents/OSS/my-mcp-server/node_modules/.bin/tsx",
      "args": ["watch", "/Users/marvelsmbp/Documents/OSS/my-mcp-server/src/index.ts"],
      "env": {
        "OPENWEATHER_API_KEY": "your_api_key_here"
      }
    }
  }
}

Claude Desktop: Add a development server entry with:

  • Command: /Users/marvelsmbp/Documents/OSS/my-mcp-server/node_modules/.bin/tsx
  • Args: ["watch", "/Users/marvelsmbp/Documents/OSS/my-mcp-server/src/index.ts"]
  • Env: OPENWEATHER_API_KEY=your_api_key_here

Development Workflow

  1. Start development server: pnpm run dev
  2. Configure MCP clients to use the tsx watch command (see configs above)
  3. Edit TypeScript files in src/ - changes auto-restart the server
  4. Test in Cursor/Claude - tools reflect changes immediately
  5. For production: Switch clients back to node dist/index.js and run pnpm run build

Testing Your MCP Server

Option 1: Direct MCP Server Usage Use the server directly with MCP clients or integrate it into other applications.

Option 2: With ADK CLI (For Client-Side Testing) While this template creates an MCP server, you can create a simple client to test it:

  1. Install the ADK CLI globally:

    npm install -g @iqai/adk-cli
    
  2. Create a simple agent that uses your MCP server (in a separate test directory):

    # In a test directory, create an agents/agent.ts file that connects to your MCP server
    adk run  # Test the integration
    adk web  # Web interface for testing
    

This approach is useful for testing how your MCP server tools work within the ADK ecosystem.

Using the Server

After building (pnpm run build), you can run the server:

  • Directly if linked or globally installed: mcp-hello-server (or your customized bin name).
  • Via node: node dist/index.js
  • Via pnpm dlx (once published): pnpm dlx your-published-package-name

推荐服务器

Baidu Map

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。

官方
精选
JavaScript
Playwright MCP Server

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。

官方
精选
TypeScript
Magic Component Platform (MCP)

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。

官方
精选
本地
TypeScript
Audiense Insights MCP Server

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。

官方
精选
本地
TypeScript
VeyraX

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。

官方
精选
本地
graphlit-mcp-server

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。

官方
精选
TypeScript
Kagi MCP Server

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。

官方
精选
Python
e2b-mcp-server

e2b-mcp-server

使用 MCP 通过 e2b 运行代码。

官方
精选
Neon MCP Server

Neon MCP Server

用于与 Neon 管理 API 和数据库交互的 MCP 服务器

官方
精选
Exa MCP Server

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。

官方
精选