MCP Calculation Server
A stateless MCP server that exposes arithmetic operations (add, subtract, multiply, divide) as tools over HTTP, enabling AI assistants to perform basic calculations conversationally.
README
MCP — Calculation Server
A stateless Model Context Protocol (MCP) server that exposes arithmetic tools over HTTP. MCP clients (Claude Desktop, VS Code Copilot, Cursor, etc.) can connect to it and invoke the tools conversationally.
Architecture
MCP Client (Claude, VS Code, etc.)
│
│ HTTP POST /mcp (JSON-RPC 2.0)
▼
Express.js server :3000
│
│ per-request lifecycle
▼
McpServer ── StreamableHTTPServerTransport
│
4 calculation tools
Each incoming request gets a fresh McpServer instance that is torn down when the response closes (stateless mode).
Tools
| Tool | Description | Inputs | Output |
|---|---|---|---|
add_two_numbers |
Add two numbers | a: number, b: number |
result: number |
subtract_two_numbers |
Subtract the second number from the first | a: number, b: number |
result: number |
multiply_two_numbers |
Multiply two numbers | a: number, b: number |
result: number |
divide_two_numbers |
Divide the first number by the second | a: number, b: number |
result: number |
divide_two_numbersreturns an error response whenbis0rather than throwing.
Prerequisites
- Node.js 18+
- npm
Getting Started
1. Install dependencies
npm install
2. Run in development mode (hot reload)
npm run dev
The server starts at http://localhost:3000/mcp.
3. Build for production
Note:
tsconfig.jsoncompiles to./build/, butpackage.json'sstartscript points todist/index.js. Fix one of them before running production mode.
# Option A — update start script to match tsconfig outDir
npm run build
node build/index.js
# Option B — update tsconfig outDir to "dist" and then run
npm run build
npm start
4. Run with Docker
docker build -t mcp-calculation .
docker run -p 3000:3000 mcp-calculation
Custom port:
docker run -p 8080:8080 -e PORT=8080 mcp-calculation
Connecting an MCP Client
VS Code (Copilot / Claude extension)
A .vscode/mcp.json is already included:
{
"servers": {
"my-mcp-server-de2625e8": {
"url": "http://localhost:3000/mcp",
"type": "http"
}
}
}
Start the server (npm run dev), then open the MCP panel in VS Code — the Calculation Server will appear automatically.
Claude Desktop
Add the following to your Claude Desktop config (claude_desktop_config.json):
{
"mcpServers": {
"calculation-server": {
"url": "http://localhost:3000/mcp",
"type": "http"
}
}
}
Any HTTP MCP client
| Setting | Value |
|---|---|
| URL | http://localhost:3000/mcp |
| Transport | HTTP (Streamable HTTP) |
| Method | POST |
| Protocol | JSON-RPC 2.0 |
Using the Tools
Once connected, ask your MCP client naturally:
What is 42 plus 58?
What is 100 divided by 4?
Multiply 7 by 8.
Subtract 15 from 100.
Direct HTTP example (curl)
# List available tools
curl -X POST http://localhost:3000/mcp \
-H "Content-Type: application/json" \
-d '{
"jsonrpc": "2.0",
"id": 1,
"method": "tools/list"
}'
# Call add_two_numbers
curl -X POST http://localhost:3000/mcp \
-H "Content-Type: application/json" \
-d '{
"jsonrpc": "2.0",
"id": 2,
"method": "tools/call",
"params": {
"name": "add_two_numbers",
"arguments": { "a": 12, "b": 30 }
}
}'
Expected response:
{
"jsonrpc": "2.0",
"id": 2,
"result": {
"content": [{ "type": "text", "text": "The sum of 12 and 30 is 42" }],
"structuredContent": { "result": 42 }
}
}
# Division by zero — returns an error response (not an HTTP error)
curl -X POST http://localhost:3000/mcp \
-H "Content-Type: application/json" \
-d '{
"jsonrpc": "2.0",
"id": 3,
"method": "tools/call",
"params": {
"name": "divide_two_numbers",
"arguments": { "a": 10, "b": 0 }
}
}'
Environment Variables
| Variable | Default | Description |
|---|---|---|
PORT |
3000 |
Port the HTTP server listens on |
Project Structure
mcp-calculation/
├── src/
│ └── index.ts # Server entry point — tools + Express setup
├── .vscode/
│ └── mcp.json # VS Code MCP client config
├── Dockerfile
├── package.json
└── tsconfig.json
Tech Stack
| Package | Purpose |
|---|---|
@modelcontextprotocol/sdk |
MCP server + transport primitives |
express |
HTTP server (v5) |
zod |
Input/output schema validation |
tsx |
TypeScript execution for dev mode |
typescript |
Type checking + build |
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。