Pulsetic RUM

Exactly something which I was looking for my product :)

"catch the slow checkout that a 200 response code hides" is a real problem, synthetic uptime checks tell you the server answered, not that a real visitor had a usable experience. the part I'd want to understand is sampling: does the async script report every real visitor's Core Web Vitals, or does it sample a percentage on high-traffic sites, and if it samples, can you slice down to a specific slow segment (one country, one browser) without losing statistical confidence? that's usually where RUM tools get uncomfortable to use in practice.

Hi Product Hunt,Andrian here. Some of you already use Pulsetic without knowing it: we run the public status trackers for Claude, ChatGPT, Cursor and Vercel. On Claude alone that is 97 incidents in the last 90 days, 32 of them major, and we usually see the disruption before the official status page posts one.We first launched Pulsetic here in 2020 and have come back five times since. Every one of those launches answered the same question: is your website up, and how do you tell customers when it isn't. This one answers a different question, and it started with a bad afternoon.A customer opened a checkout, waited four seconds for it to paint, and left. Uptime was 100%. Every check we ran returned 200. We had spent six years building monitoring that could not see the thing that actually cost the money.So Pulsetic now measures real visitors: Core Web Vitals in the field, load times at p50, p75 and p95, and the JavaScript errors people actually hit, sliced by page, country, browser and device. One async script, no cookies, nothing stored on the visitor's device. The gallery shows it better than I can describe it.RUM is the headline, but it isn't the only thing that shipped since we were last here:Real User MonitoringPulsetic AI, which reads your account and asks before it changes anythingAn MCP server, 17 tools, for Claude Code, Cursor and Windsurf (Team plan)Cron monitoring, on heartbeatsDomain expiry monitoringScreenshots captured at the moment a check failsAnnotations, so a deploy is marked on the response-time chartCloudflare allowlisting, which keeps our checker IPs on your firewallEngineers at large companies run this alongside their observability stack rather than instead of it, for the jobs that stack structurally cannot do: a status page that lives outside the infrastructure that is down, SLA reporting that stops being a manual month-end job, and the domains and scheduled jobs nobody owns until one of them fails quietly.If you already run traces and application-level instrumentation, we are not replacing that. RUM tells you what your users felt, not which query was slow.Tell me what is missing from the RUM data. We shipped a deliberately narrow first version and I would rather add the two things you would open every week than twenty you would click once. I am here all day!