mcp-stdio
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@mcp-stdiogreet Bob"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
mcp-stdio
Zero-dependency MCP server for stdio transport. Just tools, nothing else.
Why
The official MCP SDK has 17 dependencies (Express, Hono, OAuth, JWT...). Most MCP servers just need stdio + tools. This package is 222 lines, zero runtime dependencies.
|
| |
Dependencies | 17 | 0 |
Transport | stdio, HTTP, SSE | stdio |
Features | tools, resources, prompts, sampling, auth | tools |
Lines | ~5000+ | 222 |
Related MCP server: MPC-PoC-1-EchoServerToolsbasicas
Install
npm install mcp-stdioUsage
import { createMcpServer } from 'mcp-stdio';
createMcpServer({
name: 'my-server',
version: '1.0.0',
tools: {
greet: {
description: 'Say hello',
parameters: {
type: 'object',
properties: { name: { type: 'string' } },
required: ['name'],
},
handler: async ({ name }) => `Hello ${name}!`,
},
add: {
description: 'Add two numbers',
parameters: {
type: 'object',
properties: {
a: { type: 'number' },
b: { type: 'number' },
},
required: ['a', 'b'],
},
handler: async ({ a, b }) => `${Number(a) + Number(b)}`,
},
},
});Configure in Claude Code
{
"mcpServers": {
"my-server": {
"command": "node",
"args": ["my-server.js"]
}
}
}API
createMcpServer(options)
Starts an MCP server reading JSON-RPC from stdin, writing to stdout.
Options:
Field | Type | Required | Description |
|
| yes | Server name (shown to clients) |
|
| no | Server version (default: |
|
| yes | Tool definitions |
ToolDefinition:
Field | Type | Required | Description |
|
| yes | What the tool does |
|
| no | JSON Schema for input |
|
| yes | Implementation |
Handler return types:
string— wrapped as[{ type: 'text', text: '...' }]ToolContent[]— returned as-is (supportstext,image,resource)
Errors in handlers are caught and returned as { isError: true, content: [{ type: 'text', text: 'Error: ...' }] } — the server never crashes.
What it implements
JSON-RPC 2.0 over stdio (newline-delimited)
initializewith capability negotiationtools/listwith JSON Schema input schemastools/callwith structured content responsespingProper error codes (-32700, -32600, -32601, -32603)
Notification handling (no response for messages without
id)Logs to stderr (stdout is protocol-only)
What it doesn't implement
HTTP/SSE/Streamable HTTP transport
Resources, Prompts, Sampling
OAuth, authentication
Schema validation (trusts the caller — your handler validates)
These are intentional omissions. If you need them, use the official SDK.
License
MIT
This server cannot be deployed
Maintenance
Related MCP Connectors
Host your MCP tool over streamable HTTP in one command.
MCP server for progressive tool usage at any scale (see https://klavis.ai)
Related MCP Servers
- AlicenseNot gradedqualityDmaintenanceA simple toolkit for creating MCP servers with stdio and SSE transport, auto-validating tools via Pydantic.MIT
- FlicenseAqualityCmaintenanceA learning-oriented MCP server that exposes basic tools (echo, add, reverse) over stdio transport to validate MCP handshake and tool calls.3-
- AlicenseNot gradedqualityBmaintenanceA minimal, decorator-based framework for building MCP servers in Python with zero dependencies. It enables you to define tools and resources via simple decorators, handling JSON-RPC protocol, type inference, and stdio transport automatically.MIT
- AlicenseNot gradedqualityAmaintenanceMinimal MCP server allowing Python functions to be served as MCP tools over stdio or HTTP, with a built-in client for connecting to MCP servers.2Apache 2.0