Silent Outage

Monitoring denominated in dollars and dependencies, not pings.

Your uptime monitor is green. Your checkout has been broken for forty minutes.

Silent Outage watches the money and the providers behind it, not just the server. When revenue stops it pages you with a dollar range for what the outage has cost so far — and a verdict on whether the fault is yours or a provider's.

Is it you, or is it your dependency?

Start free

No card, no password, no sales call. 10 free checks, and a paid layer only when your revenue is worth watching.

Estimated revenue lost$412.00–$689.00Whose fault it is[PROVIDER] Likely provider

A real revenue alert, rendered by the same code that sends one. Figures are illustrative. Every alert carries the same eight fields — what broke, since when, what it has cost, whose fault it looks like, the evidence behind that verdict, the raw signals it was read from, and what to check first — on email, Slack and Discord alike.

Same outage. Two monitors.

2:23pm on a Monday. Your payment provider starts refusing card authorisations. Your site is up, every page returns 200, and not one customer can pay you.

  • A commodity uptime check

    Watching: does the server answer?

    • HTTP 200 from every location.
    • Response time normal.
    • No incident. No email. No page.

    You find out at 3:04pm, from a customer, in a support ticket.

  • Silent Outage

    Watching: is money still moving, and is your stack healthy?

    • Charges in the last 30 minutes: 0, against 34 ± 9 for this hour of this weekday.
    • Estimated loss so far, as a range with a confidence label — never a single false-precise figure.
    • Your provider posted a partial outage six minutes before we detected it, and our own probes of their API were already failing.
    • Verdict on the page: likely provider, likely you, or inconclusive — with the evidence it rests on.

    You are paged at 2:24pm, and the first line already tells you it is not your deploy.

A ping check answers "is the server up". That has been a free commodity for a decade, and it is free here too. The question worth paying for is whether you are still getting paid.

What you are actually buying

Three layers, and the order matters. The first two are the reason to pay; the third is a commodity we hand over so that the instrument and the product live in the same place.

  • Dollars

    An outage is priced, not just reported.

    A revenue alert carries an estimated loss range with a confidence label, computed against your own typical revenue for that hour of that weekday — never a bare "check is down", and never a single false-precise figure. Too quiet an hour to judge, and it says "below detection floor" rather than implying you lost nothing.

  • Dependencies

    Every incident says whether it is you or one of them.

    Around twenty providers an AI-era stack actually runs on, watched deeply rather than a six-thousand-logo catalogue watched shallowly. We read each one’s status feed and probe its API ourselves every sixty seconds, so a vendor page that has not caught up yet is still evidence. Every verdict is likely provider, likely you, or inconclusive — and inconclusive is an answer we are willing to give.

  • Pings

    The commodity, given away.

    Cron, heartbeat and HTTP uptime monitoring is table stakes, so it is free: ten checks, unique ping URLs, start/success/fail signals, measured run durations, and agreement from more than one location before anything alerts. It is the instrument, not the product.

The stack we watch for you

These are the providers we watch, in full, today — every one of them read from the same catalogue the monitors run against. Not a logo wall: for each one we poll its own status feed and independently probe its API, and either can put a verdict on your incident.

  • OpenAI
  • Anthropic
  • Stripe
  • Vercel
  • Supabase
  • Cloudflare
  • Twilio
  • Fly.io
  • Pinecone
  • Upstash
  • Deepgram
  • Resend
  • PlanetScale
  • ElevenLabs
  • Neon
  • AWS (incl. SES)
  • Google Cloud (incl. Gemini)
  • Railway
  • Postmark

No customer logos, no testimonials and no uptime percentage anywhere on this page: this product has not launched, and the provider list is the one piece of evidence we can show that is simply true.

Before you hand anyone a Stripe key

  • What can you do with my Stripe key?

    Read six things, and nothing else. You paste a read-only restricted key scoped to Charges, PaymentIntents, Checkout Sessions, Invoices, Payouts and Events; Customers and PaymentMethods stay at None, so the key physically cannot reach card data or your customers’ personal details. We check the scopes ourselves at connect time rather than taking your word for it, and we store amounts, counts, currency, timestamps and Stripe event ids — never a payload.

  • What happens when you go down?

    The thing that sends your alerts runs on different infrastructure from the thing that receives your pings, so an outage on our ingestion edge cannot silence a page. A third-party dead-man’s-switch watches us, and we are deliberately not in its notification path — monitoring ourselves with ourselves is worthless during exactly the outage that matters. And when a mass of checks goes quiet at once we read that as ours, mark them unknown and alert ourselves instead of paging every customer.

  • Am I locked in?

    The core is open source under the AGPL-3.0 and you can run it yourself — the hosted service is this same code, and every page carries a link to the source it is running. Instrumenting is one HTTP request per job; there is a Node package, a Python package and a curl one-liner, all three dependency-free, and none of them can throw inside your app.

How it works

Four steps, and your first monitor is live in under ten minutes. Nothing to install on a server, and nothing to run.

  1. Sign in with an emailed link

    No password to choose, store or lose. Your email address is the only personal data we hold.

  2. Paste a crontab line, or add two lines of code

    A cron expression plus a timezone and a grace time becomes a monitor with its own ping URL. Or call one URL from the job you already have.

  3. Connect Stripe with a read-only restricted key

    We learn your normal revenue for each hour of the week, then tell you when it stops. Read-only, restricted, and never card data — the scopes are checked at connect time.

  4. Pick your providers, and confirm where alerts go

    Email, Slack or Discord, with SMS and voice escalation on the paid plans. Nothing is ever sent to a destination you have not confirmed.

Plans

Every plan starts the same way, on Free. Paid plans are not self-serve yet — billing opens later and we will move you across then. Prices are per month.

  • Free

    Free

    Instrumenting a side project, or trying this before paying for it.

    10 checks

    7 days of check history

    • Heartbeat and dead-man’s-switch monitors
    • Cron monitors with a timezone and a grace time
    • HTTP uptime checks, confirmed from more than one location
    • Email alerts
    • Slack and Discord alerts, with ack and resolve from chat
    • Signed outbound webhooks on triggered, acknowledged and resolved

    Start free

  • For a solo founder

    Indie

    $19/mo

    A solo founder at a few thousand a month in revenue, with no ops and no SRE.

    More checks*

    30 days of check history

    Everything in Free, plus:

    • Revenue monitoring — Stripe webhook silence and payment-volume collapse
    • Business-event heartbeats — signups, jobs, anything you can ping us about
    • Synthetic checkout, replayed against Stripe test mode
    • Dollar-loss estimates on revenue incidents
    • Dependency monitoring for the curated provider list
    • Attribution — likely provider, likely you, or inconclusive
    • SMS escalation
    • Voice-call escalation

    Start free

  • Startup

    $49/mo

    A small team that has already had one silent checkout outage and wants the dollar figure.

    More checks*

    90 days of check history, 30 days of per-ping detail

    Everything in Indie, plus:

    • LLM quality-drift canaries: latency, refusals, output length and rubric scores

    Start free

  • Team

    $99/mo

    An agency or a team running several client apps that needs per-client separation and evidence.

    More checks*

    365 days of check history, 30 days of per-ping detail

    Everything in Startup.

    Start free

* The exact paid check allowances are still being set, so no number is quoted for them here rather than a number we might have to take back. Free is fixed at ten.