MCP Server
A unified Model Context Protocol server with pluggable tools and independent endpoints, allowing clients to selectively add specific tools like calculator, echo, and time utilities with fault isolation and secure authentication.
README
MCP Server
A unified MCP (Model Context Protocol) server with pluggable tools. Each tool has an independent endpoint, allowing clients to selectively add specific tools.
Features
- Unified Endpoint with Path Routing: Each tool has its own MCP endpoint (
/mcp/{toolName}) - Loosely Coupled Tools: Tools are independently loaded from
src/tools/directory - Fault Isolation: Individual tool failures don't affect other tools (timeout protection + error handling)
- StreamableHTTP Transport: Modern MCP transport protocol
- Bearer Token Authentication: Secure API access
- Docker Support: Ready for containerized deployment
Project Structure
mcp_server/
├── src/
│ ├── index.ts # Entry point
│ ├── core/
│ │ ├── tool-registry.ts # Tool registry
│ │ ├── tool-loader.ts # Dynamic tool loader
│ │ └── tool-executor.ts # Isolated executor (timeout + error handling)
│ ├── server/
│ │ ├── app.ts # Express + StreamableHTTP
│ │ └── mcp-server.ts # MCP Server instance
│ ├── middleware/
│ │ └── auth.ts # Bearer token authentication
│ ├── tools/ # Loosely coupled tools directory
│ │ ├── calculator/index.ts # Calculator tool (add/subtract/multiply/divide)
│ │ ├── echo/index.ts # Echo tool (echo/reverse/info)
│ │ └── time/index.ts # Time tool (get_current_time/convert_time/format_time)
│ ├── types/mcp.ts # Type definitions
│ └── utils/logger.ts # Terminal logger
├── Dockerfile # Multi-stage build
├── docker-compose.yml
├── package.json
└── tsconfig.json
Quick Start
Local Development
# Install dependencies
npm install
# Create .env file
echo "AUTHORIZATION_KEY=your-secret-key" > .env
# Start development server
npm run dev
Docker Deployment
# Build and run with docker-compose
docker-compose up -d --build
# Or build manually
docker build -t mcp-server .
docker run -e AUTHORIZATION_KEY=your-secret-key -p 3000:3000 mcp-server
API Endpoints
| Endpoint | Method | Description |
|---|---|---|
/mcp/{toolName} |
POST | MCP JSON-RPC request for specific tool |
/mcp/{toolName} |
GET | SSE connection (requires session-id) |
/mcp/{toolName} |
DELETE | Close session |
/health |
GET | Health check |
/tools |
GET | List all available tools and their endpoints |
Available Tools
Calculator (/mcp/calculator)
Math expression evaluator supporting various operators and functions.
| Method | Description | Parameters |
|---|---|---|
evaluate |
Evaluate a math expression | expression: string |
Supported Operators:
- Arithmetic:
+,-,*,/ - Floor Division:
// - Modulo:
% - Power:
**or^
Supported Constants: pi, e
Supported Functions: sin, cos, tan, sqrt, abs, log, log10, exp, floor, ceil, round, pow, min, max
Examples:
2 + 3 * 4 → 14
(2 + 3) * 4 → 20
2^10 → 1024
sqrt(16) → 4
2 * pi → 6.283...
sin(0) → 0
pow(2, 8) → 256
17 // 5 → 3
17 % 5 → 2
Echo (/mcp/echo)
Echo tool for testing and debugging.
| Method | Description | Parameters |
|---|---|---|
echo |
Echo back the input message | message: string |
reverse |
Reverse the input string | text: string |
info |
Get server information | - |
Time (/mcp/time)
Time tool for timezone conversion and formatting.
| Method | Description | Parameters |
|---|---|---|
get_current_time |
Get current time in a specific timezone | timezone: string (default: "Etc/UTC") |
convert_time |
Convert time between timezones | source_timezone: string, target_timezone: string, time: string (HH:MM) |
format_time |
Convert timestamp to formatted string | timestamp: number (ms), timezone: string |
Client Configuration
Claude Desktop
Edit ~/Library/Application Support/Claude/claude_desktop_config.json (macOS):
{
"mcpServers": {
"calculator": {
"url": "http://localhost:3000/mcp/calculator",
"transport": "streamable-http",
"headers": {
"Authorization": "Bearer your-secret-key"
}
},
"time": {
"url": "http://localhost:3000/mcp/time",
"transport": "streamable-http",
"headers": {
"Authorization": "Bearer your-secret-key"
}
}
}
}
Cursor / VS Code
Create .cursor/mcp.json in project root:
{
"mcpServers": {
"calculator": {
"url": "http://localhost:3000/mcp/calculator",
"transport": "streamable-http",
"headers": {
"Authorization": "Bearer your-secret-key"
}
}
}
}
Programmatic Usage (Node.js)
import { Client } from "@modelcontextprotocol/sdk/client/index.js";
import { StreamableHTTPClientTransport } from "@modelcontextprotocol/sdk/client/streamableHttp.js";
const client = new Client({
name: "my-client",
version: "1.0.0",
});
const transport = new StreamableHTTPClientTransport(
new URL("http://localhost:3000/mcp/calculator"),
{
requestInit: {
headers: {
Authorization: "Bearer your-secret-key",
},
},
}
);
await client.connect(transport);
// List available tools
const tools = await client.listTools();
console.log(tools);
// Call a tool
const result = await client.callTool({
name: "add",
arguments: { a: 10, b: 5 },
});
console.log(result);
Adding New Tools
- Create a new directory under
src/tools/:
src/tools/my-tool/
└── index.ts
- Implement the
MCPToolinterface:
import { z } from "zod";
import { MCPTool, MCPMethodDefinition } from "../../types/mcp.js";
const myTool: MCPTool = {
name: "my-tool",
description: "Description of my tool",
version: "1.0.0",
getMethods(): MCPMethodDefinition[] {
return [
{
name: "my-method",
description: "Description of my method",
inputSchema: {
param1: z.string().describe("Parameter description"),
},
handler: async (params) => {
const { param1 } = params as { param1: string };
return { result: param1 };
},
},
];
},
async initialize() {
// Initialization logic
},
async healthCheck() {
return true;
},
};
export default myTool;
- The tool will be automatically loaded on server startup.
Scripts
npm run dev # Start development server with hot reload
npm run build # Build for production
npm start # Start production server
npm test # Run tests
npm run lint # Check code style
npm run lint:fix # Auto-fix code style issues
License
MIT
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。