One problem · bounded diagnostic

MCP server not responding: locate the timeout

Separate a server startup failure from an initialization failure and a slow tool call before changing timeouts.

Run the free diagnostic · See service scope and current price

No payment required for the free check. A paid repair requires a separate quote and a verified available payment method.

Check startup first

Confirm that the server process stays running. For stdio transports, diagnostic logs belong on stderr: ordinary text on stdout can corrupt protocol messages. Keep a sanitized startup exit code.

Identify the failing boundary

Record whether the failure happens during initialization, tool listing or one specific tool call. A successful tools/list result narrows the problem to invocation or the downstream dependency.

Test one bounded call

Use one small known input and one retry at most. Note elapsed time and whether a response eventually arrives. Increasing a timeout does not fix a crashed process or an unreachable dependency.

Evidence to keep

Transport type, failing method, elapsed time and a sanitized error message. Do not send credentials or private endpoints.

Scope of this check

The free probe classifies the symptom. It does not connect to your server or prove that a downstream dependency is healthy.

MCP server timeout: identify the first failing protocol stage

A timeout can happen before initialize, while listing tools, or during one tools/call. Capture each stage with a sanitized status and elapsed time. This synthetic example illustrates diagnosis only: it does not contact an MCP server, execute a paid repair, or prove a live endpoint is healthy.

const stages=[
 {name:'initialize',ok:true,elapsedMs:120},
 {name:'tools/list',ok:true,elapsedMs:95},
 {name:'tools/call',ok:false,elapsedMs:30000,error:'TIMEOUT'}
];
const first=stages.find(x=>!x.ok);
console.log(first.name+': '+first.error+' after '+first.elapsedMs+'ms');

Expected output: tools/call: TIMEOUT after 30000ms. This narrows the failure to invocation or a downstream dependency, not the initialization handshake. In a real incident, validate the actual client transport, logs, and one authorized bounded test; never paste credentials or run repeated production calls.

Need a scoped review? See the service scope and current price. Accept a quote only after reviewing its scope and available payment methods.

Related troubleshooting guides

Try the free probe

Use a sanitized symptom or a synthetic sample. The request goes to the existing NAIF diagnostic API. No payment or repair is started.

Review this service, current price and quote options

Availability is determined by the existing service and country policy. A diagnosis or quote does not confirm payment or guarantee a repair.

Use this guide with an AI agent
{
  "source": "discovery_pages_v1",
  "version": "2.0.0",
  "intentManifest": "/.well-known/naif-intents.json",
  "agentCard": "/.well-known/agent-card.json",
  "page_id": "mcp-server-not-responding",
  "intent": "mcp_transport_timeout",
  "productId": "mcp-tool-mini",
  "capability": "/api/agent/catalog",
  "pricing": "/api/agent/catalog",
  "probe": {
    "method": "POST",
    "endpoint": "/api/agent/probe",
    "body": {
      "productId": "mcp-tool-mini",
      "source": "discovery_pages_v1",
      "input": {
        "message": "MCP initialization succeeds and tools/list works, but one tool times out after 30 seconds."
      }
    }
  },
  "quote": "/api/agent/quote",
  "acceptQuote": "/api/quote/accept",
  "offers": "/.well-known/naif-offers.json",
  "checkout": "Use only the checkoutEndpoint returned by the existing quote acceptance API when checkoutEligible is true.",
  "attribution": "Retain the __Host-naif_discovery cookie, or echo X-Naif-Discovery-ID returned by this page on same-origin API requests. Do not treat the identifier as authentication."
}
Existing agent discovery