Health Probes

A health probe is a lightweight request MCPHub sends to your MCP server to check if it's responding correctly.

What a probe does

MCPHub sends an MCP initialize JSON-RPC call to the server endpoint:

{
  "jsonrpc": "2.0",
  "method": "initialize",
  "params": {
    "protocolVersion": "2024-11-05",
    "capabilities": {},
    "clientInfo": {"name": "mcphub-prober", "version": "1.0"}
  },
  "id": 1
}

The probe records:

• Latency — time from request to first response byte (ms)
• HTTP status — 2xx is considered healthy
• Error — any connection error or non-2xx response

If the server requires auth, the probe uses the configured credentials automatically.

Status thresholds

Triggering probes

On demand — single server:

1. Open the server detail page
2. Click Probe Now
3. Status updates within a few seconds

On demand — all servers:

1. Go to the Servers page
2. Click Run All Probes
3. All servers are probed in parallel; status updates as results come in

Automated (once/day): MCPHub's Vercel Cron runs daily at 02:00 UTC and probes all servers across all workspaces.

Probe history

Every probe result is stored in the health_checks table. You can see the last 30 days of results in the server detail page's health timeline.