octo-mcp-server
A production-ready MCP server built with Node.js and Express that supports remote deployment via HTTP and SSE. It provides a modular framework for building and scaling tools while serving multiple clients concurrently.
README
octo-mcp-server
A production-ready Model Context Protocol (MCP) server built with Node.js, TypeScript, and Express. It uses HTTP + SSE (Server-Sent Events) as the transport so it can be deployed remotely and serve multiple clients concurrently.
Project structure
src/
index.ts # Entry point — creates Express app, mounts /sse and /message routes
server.ts # MCP server factory, session registry, message routing
tools/
echo.ts # Example tool: returns its input unchanged
Dockerfile
tsconfig.json
package.json
.env.example
Adding a new tool
- Create
src/tools/my-tool.tsand export aregisterMyTool(server: McpServer)function:
import { McpServer } from '@modelcontextprotocol/sdk/server/mcp.js';
import * as z from 'zod/v4';
export function registerMyTool(server: McpServer): void {
server.tool(
'my-tool',
'Description shown to the LLM',
{ input: z.string() },
async ({ input }) => ({
content: [{ type: 'text', text: `Result: ${input}` }],
}),
);
}
- Import and call it in
src/server.ts:
import { registerMyTool } from './tools/my-tool.js';
// inside createMcpServer:
registerMyTool(server);
That's it.
Local development
Prerequisites
- Node.js ≥ 18
- npm
Setup
cp .env.example .env # edit PORT or add API keys as needed
npm install
npm run build
npm start
The server starts on http://localhost:3000 (or whatever PORT is set to).
Verify with curl
# Health check
curl http://localhost:3000/health
# Open an SSE stream (keep this terminal open)
curl -N http://localhost:3000/sse
# You will see: event: endpoint\ndata: /message?sessionId=<ID>
# In another terminal, call the echo tool (replace <SESSION_ID>)
curl -X POST "http://localhost:3000/message?sessionId=<SESSION_ID>" \
-H "Content-Type: application/json" \
-d '{
"jsonrpc": "2.0",
"id": 1,
"method": "tools/call",
"params": {
"name": "echo",
"arguments": { "message": "hello world" }
}
}'
Available npm scripts
| Script | Description |
|---|---|
npm run build |
Compile TypeScript → dist/ |
npm start |
Run the compiled server |
npm run dev |
Run with ts-node (no compile) |
npm run clean |
Delete dist/ |
Docker
# Build
docker build -t octo-mcp-server .
# Run (pass environment variables with -e or --env-file)
docker run -p 3000:3000 --env-file .env octo-mcp-server
Deploy to Azure Web App
Option A — Deploy a Docker container (recommended)
- Push your image to a container registry (Azure Container Registry, Docker Hub, etc.):
az acr build --registry <your-registry> --image octo-mcp-server:latest .
- Create the Web App (Linux, Docker):
az group create --name rg-octo-mcp --location eastus
az appservice plan create \
--name plan-octo-mcp \
--resource-group rg-octo-mcp \
--is-linux \
--sku B1
az webapp create \
--name octo-mcp-server \
--resource-group rg-octo-mcp \
--plan plan-octo-mcp \
--deployment-container-image-name <your-registry>.azurecr.io/octo-mcp-server:latest
- Set environment variables:
az webapp config appsettings set \
--name octo-mcp-server \
--resource-group rg-octo-mcp \
--settings PORT=3000 MY_API_KEY=...
- Enable Always On so the server stays warm:
az webapp config set \
--name octo-mcp-server \
--resource-group rg-octo-mcp \
--always-on true
SSE note: Azure App Service by default has a 230-second idle request timeout. For long-lived SSE connections you may need to send keep-alive comments or configure the timeout via
az webapp config set --http20-enabled true.
Option B — Deploy from source with Oryx build
- Zip the source (exclude
node_modulesanddist):
zip -r deploy.zip . --exclude "node_modules/*" "dist/*" ".git/*"
- Deploy:
az webapp deployment source config-zip \
--name octo-mcp-server \
--resource-group rg-octo-mcp \
--src deploy.zip
Azure will run npm install && npm run build && npm start automatically via the Oryx build system.
Environment variables
| Variable | Default | Description |
|---|---|---|
PORT |
3000 |
TCP port the server listens on |
Add any tool-specific secrets (API keys, connection strings) here as well.
API endpoints
| Method | Path | Description |
|---|---|---|
| GET | /health |
Returns {"status":"ok","timestamp":"…"} |
| GET | /sse |
Opens an SSE stream; emits the POST endpoint + session ID |
| POST | /message |
Receives JSON-RPC 2.0 messages (?sessionId=<id>) |
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。