know it’s running.
Operations is Fizzgig watching the things you can’t watch continuously. Availability, deploys, spend, and the anomalies that only look like anomalies if someone is comparing today against last week. It pulls from every platform your product runs on and reduces the result to one ranked list - so the question “is anything wrong?” has an answer you can get in ten seconds rather than twelve browser tabs.
nothing is broken. probably.
Your product runs on six platforms and each one has a dashboard. Each dashboard is fine. Each one shows you its own slice, in its own layout, with its own idea of what counts as urgent. None of them tells you about the other five.
So the things that go wrong are the things that fall between them. The deploy that succeeded on Vercel while the migration it depended on failed on Supabase. The security advisor that appeared on Tuesday and will still be there in September. The spend that crept from £40 to £180 across three platforms, none of which individually looked alarming. The overnight failure you found out about from a user.
And the honest constraint: you are one person, or four, and you’re shipping. Nobody is going to check six dashboards every morning. It doesn’t matter how important it is - it isn’t going to happen.
So Fizzgig checks them. Every one, on a schedule, and tells you only what changed.
one surface, four kinds of signal.
uptime_statussoonAvailability and response-time checks against every surface you’ve deployed.
Scheduled probes per deployed endpoint, cross-referenced with deploy events from your hosting platforms so a blip during a deploy reads differently to a blip at 3am.
Downtime you hear about from Fizzgig rather than from a customer. Latency creep, which is the more common and more insidious failure - nothing is down, everything is slower than it was a fortnight ago, and no single day’s change is big enough to notice.
triage_listsoonCollapses every platform’s notifications into one ranked list.
Pulled from all connected platforms on your plan’s refresh schedule, then summarised by Claude Haiku into a triage order rather than a chronological pile.
Failed deploys, unread PR reviews, critical dependency alerts and security advisor errors in one place, ordered by what actually needs you. The ranking is the product here - an unranked aggregation of six platforms’ notifications is just a longer list.
cost_summarysoonTracks what your product costs across every platform, and where that’s heading.
Plan and usage data from each connected platform, combined into month-to-date spend and a projection.
One number for what you’re spending, broken down by platform. Plan estimate against observed usage, so you can see which platform you’ve outgrown and which you’re paying for twice. Warnings as you approach caps - before the overage, which is the only point at which a warning is any use.
anomaly_scansoonNotices when a number stops behaving the way it has been.
Compares current metrics against your own recent baseline, per platform.
Deploy-failure rate spikes, unexpected database growth, new ERROR-level advisors, revenue drops. The value is the baseline: a 40% jump in failed deploys matters, and no dashboard will tell you it happened because every dashboard shows you the absolute number rather than the change.
your stack, every morning.
Overnight, Fizzgig checks every connected platform. By 07:30 UTC there’s one page in your inbox - and if nothing needs you, no email fires at all. Silence is the default; noise has to earn its way in.
Two things on that page you would not have found yourself: an advisor warning on a table you weren’t looking at, and 90 milliseconds of latency that arrived gradually enough to be invisible.
what operations gives you as you grow.
Daily refresh and a digest, so a hobby project can’t quietly rack up a bill or sit broken for a fortnight while you’re not looking.
The rung where this pillar earns its keep. Real users mean uptime matters, real deploys mean failures matter, and real spend means the Monday morning bill is a genuine risk. Hourly refresh, cost caps, MRR against burn.
A triage surface the whole team works from rather than one founder’s inbox. Fifteen-minute refresh across up to thirty platforms and multiple products.
Every live integration feeds operations. Vercel, Netlify and Railway for deploys and hosting. GitHub and GitLab for reviews and workflow runs. Supabase and Neon for database health and advisors. Stripe for revenue. All eight report into the same triage surface.
One token per platform, no agents to install, nothing running in your infrastructure.