Free tool
Is your PaaS down, or is it you?
I probe seven hosting platforms every minute: their API, and their own status page. Each row shows whether the API answers right now, and 90 days of history.
Vercel
LoadingRailway
LoadingRender
LoadingFly.io
LoadingSupabase
LoadingNetlify
LoadingHeroku
Loading
How the probe works
Every 60 seconds a probe checks each platform twice. It calls the platform's public API to see whether it answers at all. It reads the platform's status page to see what the platform says about itself.
The two can disagree. People write a status page, and an incident often reaches it minutes after the API stops answering. The API probe needs nobody to write anything.
A probe fails when the API gives no answer within 10 seconds, or answers with a 5xx status. Two failed probes in a row open an incident. Two good probes in a row close it, and the page lists how long it lasted. Two outage reports in a row from a status page open an incident too, marked as reported.
What gets probed
| Platform | API probed | Status page read |
|---|---|---|
| Vercel | api.vercel.com/v2/user | vercel-status.com |
| Railway | backboard.railway.com/graphql/v2 | status.railway.com |
| Render | api.render.com/v1/services | status.render.com |
| Fly.io | api.machines.dev/v1/apps | status.flyio.net |
| Supabase | api.supabase.com/v1/projects | status.supabase.com |
| Netlify | api.netlify.com/api/v1 | netlifystatus.com |
| Heroku | api.heroku.com/account | status.heroku.com |
The probe sends no credentials, so an answer like 401 Unauthorized counts as the API answering. It follows no redirects.
Reading a row
The badge is the state right now. Answering means the API answered the last probe, no incident is open, and the status page reports no outage. When the status page could not be read, the row says so. Last probe failed means one probe got no answer, which is not yet an incident. Not answering means an incident is open and the API still gives no answer. Recovering means the API answers again, and the incident closes after two good probes. Outage reported means the status page says so, while the API may still answer.
Each of the 90 bars is one UTC day. Green means the API answered every probe that day. Amber means a failed probe or a status-page report that opened no incident. Red means an incident was open that day. A low grey line means no data.
What it can't tell you
The probe runs from one place, Railway's Virginia region. It can miss an outage limited to another region. A platform's API can answer while your app's region, build queue or database is down, and the reverse.
Because the probe itself runs on Railway, a Railway failure in Virginia can stop the probe. Then every row says No recent probe, not only Railway's.
Questions
Is Railway down right now?
The Railway row answers that. Its badge says Not answering when Railway's API fails two probes in a row, and Outage reported when Railway's status page reports one. Today's bar turns red when an incident opens: two failed API probes in a row, or two outage reports in a row. A single failure or report leaves the day amber. A row that says No recent probe has had no probe for 3 minutes. The probe runs on Railway, so a Railway failure in Virginia can stop every row updating.
Is Vercel down, or is it my deployment?
When the Vercel row says Answering, Vercel's API answered the last probe, no incident is open, and its status page reports no outage. A failing site then usually points at your own deployment, a DNS record or one Vercel region. The probe runs from one place, Virginia, so an outage limited to another region can miss it.
Where can I check Supabase status?
Supabase's own page is status.supabase.com. The Supabase row here shows what that page says, plus whether api.supabase.com answered the last probe. A project can still fail while both are fine, since each project runs its own database.
What counts as down?
A probe fails when the API gives no answer within 10 seconds, or answers with a 5xx status. Two failed probes in a row open an incident, and two good ones close it. A 401 or 404 counts as an answer, because the probe sends no credentials.
How long is history kept?
90 days. Each bar is one UTC day, and each day shows how many probes the API answered. Incidents list when they started and how long they lasted.
Run your app on multiple servers without Kubernetes
See how it worksFacts checked 2026-10-17.