Test your MCP client with the official reference 'everything' server

Recipe: MCP client smoke-test via @modelcontextprotocol/server-everything

Package: @modelcontextprotocol/server-everything v2026.1.26 (npm) Launch: npx -y @modelcontextprotocol/server-everything Transport: stdio (also supports SSE and Streamable HTTP) Auth: none — fully credential-free

What it exercises

The "everything" server is the official MCP reference implementation. It exposes every protocol capability in one process, making it the ideal target for validating MCP client code.

Tools (13)

Resources (7 static markdown docs)

demo://resource/static/document/{architecture,extension,features,how-it-works,instructions,startup,structure}.md

Prompts (4)

Recipe: complete client validation

import { Client } from '@modelcontextprotocol/sdk/client/index.js'
import { StdioClientTransport } from '@modelcontextprotocol/sdk/client/stdio.js'

const transport = new StdioClientTransport({
  command: 'npx',
  args: ['-y', '@modelcontextprotocol/server-everything']
})
const client = new Client({ name: 'my-client', version: '1.0.0' }, { capabilities: {} })
await client.connect(transport) // ~3800ms cold start (npx install)

// ── Tools ──
const { tools } = await client.listTools()            // 13 tools
await client.callTool({ name: 'echo', arguments: { message: 'ping' } })
// → "Echo: ping"

await client.callTool({ name: 'get-sum', arguments: { a: 42, b: 17 } })
// → "The sum of 42 and 17 is 59."

await client.callTool({ name: 'get-tiny-image', arguments: {} })
// → [text, image(png, 5380 bytes base64), text]

await client.callTool({
  name: 'trigger-long-running-operation',
  arguments: { duration: 1, steps: 2 }
})
// → completes in ~1000ms with progress notifications

// ── Resources ──
const { resources } = await client.listResources()     // 7 markdown docs
await client.readResource({ uri: 'demo://resource/static/document/architecture.md' })
// → full markdown text with architecture overview

// ── Prompts ──
const { prompts } = await client.listPrompts()         // 4 prompts
await client.getPrompt({ name: 'simple-prompt', arguments: {} })
// → 1 message

await client.close()

Performance (measured)

• Cold start (initialize): 3801ms (npx resolution + TypeScript compile)
• tools/list: 7ms
• resources/list, prompts/list: 1ms each
• echo, get-sum, get-tiny-image: 1–3ms each
• trigger-long-running-operation(1s, 2 steps): 1006ms (expected — honors duration param)
• resources/read: 1ms

Gotchas

• Cold start is ~4s due to npx + TypeScript compilation. Warm runs (package cached) are ~1s.
• Tool names use kebab-case (get-sum, not add; get-tiny-image, not getTinyImage).
• get-structured-content requires a city enum argument: 'New York' | 'Chicago' | 'Los Angeles' — call