- 15s
- Check interval
- 11s
- Median detection
- 99.98%
- Our own uptime, last year
- 3
- False alerts last month
We find out in eleven seconds and tell you in fourteen
Dithy checks your endpoints from six locations every fifteen seconds, confirms a failure from a second region before it wakes anybody, and sends one alert rather than forty. Our own numbers are below and they are the live ones.
Pricing
See everythingWhat happens when an endpoint fails
-
1
Second 0 — a check fails
A single failing check from a single region is not an outage, it is a bad route. Nothing is sent and nothing is logged as an incident.
-
2
Second 4 — a second region confirms
The same endpoint is checked immediately from a different continent. If that one succeeds, the first result is discarded and recorded as a network blip rather than as downtime.
-
3
Second 11 — it is declared down
Two regions failing is an incident. The clock on your status page starts here, not at the first failed check, so the number you publish is the number that was real.
-
4
Second 14 — one alert, one channel
To the on-call person for that monitor, once, on their chosen channel. Everyone else gets a digest. This is the reason our users report three false alerts a month rather than three hundred.
The hard part is not noticing. It is not crying wolf
Any tool can curl a URL every minute. What separates them is what happens at three in the morning when a route flaps for nine seconds.
Every monitoring product we have used defaults to alerting on a single failed check, because it makes the detection number in the marketing look good. It also means an on-call engineer is woken by transit noise twice a week, and after a month nobody reads the alerts. Dithy requires confirmation from a second region and groups everything downstream of a failed dependency into one incident. The detection number is eleven seconds instead of four. The alerts get read.
What teams say
Forty alerts a week became three a month and all three were real.
- They publish their own post-mortems. Our last vendor showed green during an outage.
- Free tier for two years, nothing removed when I finally paid.
- Detection is eleven seconds, not four. They explain why and I agree with it.
- Unlimited team members on every tier including free.
- CSV export of raw checks without opening a support ticket.
- Private regions cost nearly double. Still cheaper than the alternative.
Questions engineers actually ask
Search the site-
Where do the checks run from?
Six regions: Virginia, Oregon, Frankfurt, London, Singapore and São Paulo. A monitor is checked from all six on rotation and a failure is always confirmed from a different continent before anything is sent. -
What counts as downtime?
Two regions failing the same check within four seconds of each other. The clock starts at that confirmation, not at the first failed check, and it stops at the first successful one. The definition is the same on your status page as in your dashboard. -
How many alerts will I get?
One per incident, to the on-call person for that monitor. Everything downstream of a failed dependency is grouped into the same incident rather than alerting separately. Teams report about three false alerts a month. -
What happens when Dithy is down?
Alerting runs in a separate path from the dashboard, so the dashboard can be down while alerts keep flowing. When both go, you get a message from the backup path and a public post-mortem inside a week. Both of last year’s are already published. -
Do you charge per user?
No. Team members are unlimited and free on every tier including the free one. Pricing is per monitor, monthly, with no annual commitment and no enterprise tier that unlocks features. -
Can you check things behind a firewall?
Yes, with a private region — a small agent inside your network that runs the same checks and reports out. It is eleven dollars per monitor rather than six and there is no separate platform fee. -
How do I leave?
Cancel in the dashboard, export every raw check result as CSV without asking us, and the account drops to the free tier rather than being deleted. We do not have a retention team and nobody will telephone you.
Point it at one endpoint and see what happens tonight
No card, no trial clock, no sales call. Ten monitors at a one-minute interval stay free for as long as you want them, and nothing you set up on the free tier is taken away if you never pay.
Incidents and changes
EverythingTalk to us
Setup questions, a migration from something else, or an enterprise volume. A real engineer replies, usually within the working day.
- 08:00 to 18:00
- 09:00 to 13:00
- Closed
Kingston, Kingston