Blog

1. AI agent uptime monitoring from outside

   An agent fails when its model API, MCP servers, or sandbox provider fails. Which endpoints to check from outside, how to check an MCP server, and what to skip.

   Read full article

2. Kubernetes ingress monitoring from outside

   Kubernetes probes never see your ingress, DNS, or load balancer. Check /readyz from outside the cluster, assert the body, and alert only on confirmed failures.

   Read full article

3. Add the check when you add the endpoint

   The health check usually gets added after the first outage. Ask your coding agent to add it in the same session it writes the endpoint, over MCP.

   Read full article

4. TTS latency monitoring: slow or down?

   Your text-to-speech API can be up and still too slow for a voice agent. How to watch TTS uptime and latency from outside, and what to alert on.

   Read full article

5. Only some pages return 502

   Homepage is 200 but one path returns 502? Curl both and read the status lines: the proxy answered, and that path's upstream did not.

   Read full article

6. Too many redirects

   Print the redirect chain and stop when a URL repeats. One hop from HTTP to HTTPS is normal. A loop never returns a status code.

   Read full article

7. DNS works, the site does not

   dig returned an address and the page still does not load. Test the port, then the certificate, then the HTTP status, and stop at the first failure.

   Read full article

8. When 403 means up

   A WAF 403 is not an outage. A 200 with an error payload is. Scan HTTP status codes for website and API checks, and what to set as expected.

   Read full article

9. Why we ping from different regions

   A check from one city only answers “is it down for me?” Multi-region pings separate a global outage from a regional routing problem.

   Read full article