Alerts that reach you and do not repeat.

The rules

A Mepa8 alert is a message that a number crossed a line you set or a vendor reported a state that will stop your product. There are 4 rules, and every vendor connector maps its own fields onto them so the alerts read the same whichever vendor fired.

Rule	Warning	Critical	What the message carries
Budget pace	On-pace figure above the monthly budget	Above 150% of it	Vendor, pace, budget, the product or table driving it where the API says
Limit	80% of a plan limit used	100% used, and whether overage billing is on	Used, limit, reset date
Past due	—	Any status other than active, a paused workspace, a failed card	The vendor's own status text and the fix
Trial and deploy	Trial ends inside 7 days; main 6 hours ahead of production	Trial ended; main 48 hours ahead	Days left, hours behind, likely cause

Every message ends with an action line, written when the rule was written, not generated on the fly. “Upgrade tier or cut non-essential TTS until the reset” is an action line. “Usage is high” is not, and Mepa8 never sends one like it.

Channels and levels

• Email is on from sign-in, to the address you signed in with. Both levels go there.
• SMS is for critical only by default. As of October 9, 2026 SMS is marked pending on the Alerts page: it ships the day a real text to a real phone is proven from the hosted app, and this page's change log will say so.
• Quiet hours apply to warnings, never to critical. Set them on the Alerts page in your local time. A warning that fires at 2 a.m. waits until the window opens; a critical one does not wait.

Why a repeat becomes one incident

On September 13 and 14, 2026 our own earlier monitor sent 14 identical critical emails in 24 hours about an ElevenLabs past-due status, into an inbox nobody was reading, while every phone agent was refusing calls. The 14th message was no more useful than the 1st. A separate alert about a stuck onboarding record had fired every 35 minutes since May. That is how a team learns to ignore alerts.

So Mepa8 opens one incident per vendor per rule. The first message goes out. The same condition on the next check updates the incident's last-seen time and count instead of sending again. When the condition clears, one resolution message goes out with how long it lasted. An incident nobody acknowledges within 7 days is surfaced on the board as stale, because an alert nobody acts on is a bug in the alert, not a fact about the vendor.

“An alert is a promise to act. If no one will act on it, it should not exist.”Mepa8 blueprint, 2026-10-08

Test it before you need it

Press Test on the Alerts page. Mepa8 sends a real message with a clearly labelled fake critical event to every channel you have, within 1 minute. If a channel does not receive it, fix the channel now; the first real alert is a bad time to learn the number was wrong. The test is recorded in the audit log so you can show, later, that the channel worked on a given date.

The reason this matters is in the numbers vendors publish about their own limits. ElevenLabs, for example, keeps answering past the credit limit only when usage-based billing is on, per its pricing page; with it off, the API stops. The alert tells you which of those two you are facing, and that is only useful if it arrives.

Questions people ask

Straight answers

How fast does an alert arrive after the condition?

Checks run every hour on the free tier, so up to 60 minutes after the vendor's API first reports the condition, plus the vendor's own reporting lag. Email sends within seconds of the check; the Alerts page shows the last check time so you know the window.

Can I get SMS alerts today?

Not yet, as of October 9, 2026. SMS is built behind a flag and ships the day a real text to a real phone is proven from the hosted app. Until then every alert goes to email, and the Alerts page states the channel status in plain words.

Will I get the same alert every hour?

No. The same condition updates one open incident with a count and a last-seen time. You get the first message, then one resolution message when it clears. If 14 identical messages in a day were useful, we would have kept the old monitor.

Sources

1. ElevenLabs — pricing (usage-based billing behaviour) — read 2026-10-08
2. Railway — usage limits (soft and hard limit notifications) — read 2026-10-09

Change log: October 9, 2026 — First published. SMS marked as pending proof.