Isobath site (isobath.io)
- 24h
- 100% (42/42)
- 7d
- 100% (42/42)
- 30d
- 100% (42/42)
- p50 latency
- 46ms 24h
Status
All systems operational
Measured every minute by an on-box probe hitting the origin, bypassing Cloudflare — this is how reliable the server actually is, not how it looks through the edge cache.
MCP server (isobath.io/mcp)
- 24h
- 100% (42/42)
- 7d
- 100% (42/42)
- 30d
- 100% (42/42)
- p50 latency
- 14ms 24h
How this is measured
A systemd timer on the same host as the app runs scripts/probe-uptime.js every 60 seconds. Each run hits /healthz and /mcp directly on the loopback interface, times the response, and writes a row to uptime_probes. A 2xx within 5 seconds is "ok"; anything else — 4xx, 5xx, timeout, network error — is "not ok". No external pingers, no synthetic monitoring vendor, just the server reporting its own liveness.
If this page itself is down, so is everything else on the box. Treat an unreachable /status as the real signal.