Setup, start to finish
- In Slack, create an incoming webhook for the channel you want (Slack’s own feature, found in your workspace’s app settings). You get one URL.
- In your Weckr project’s alert settings, paste the webhook URL and set your thresholds: the velocity window and the margin threshold.
- Hit the test alert button. A sample alert lands in the channel within seconds, delivery confirmed before any real incident depends on it.
The trust model is deliberate: the webhook is scoped to one channel, lives in your Slack admin, and can be revoked there without touching Weckr. For a product whose whole security posture is minimizing what it can see and do (covered here), asking for workspace OAuth to post messages would be backwards.
What actually arrives in the channel
- Velocity alerts: a user crossed the burn threshold, default 50,000 tokens in 5 minutes, which is the stuck agent signature. The full detection flow is in the runaway article; the alert names the user and the burn so you can act on it directly.
- Margin alerts:a paying user’s margin hit your threshold, at or below breakeven, the quiet problem from finding your first unprofitable user surfaced without you checking the dashboard.
- Weekly digest:Monday morning email with the week’s spend, top users, and top features, the ambient awareness layer under the incident alerts.
Alerts inform, caps guarantee
An alert is only as good as whoever reads it, which is fine at 2pm and useless at 2am. That is why the alert layer pairs with spending caps: the alert tells you a user is burning, the cap ensures the burn has a ceiling whether or not anyone is awake. Both read the same per user stream from the same two line integration, so adding the second once you have the first is a dashboard setting, not new code.
FAQ
How do I get AI cost alerts in Slack?
With Weckr: create an incoming webhook in your Slack workspace (a built in Slack feature, one URL), paste it into your project’s alert settings, and hit the test button to confirm delivery. From then on, runaway usage alerts and margin alerts post to that channel as they fire. No Slack OAuth, no app install from our side, and the webhook URL stays in your control.
What kinds of alerts does Weckr send?
Three: velocity alerts when a user burns tokens abnormally fast (default 50,000 tokens in 5 minutes, the runaway agent signature), margin alerts when a paying user’s margin drops to or below your threshold, and a weekly digest email every Monday summarizing spend, top users, and top features. Velocity and margin alerts deliver to Slack and email; thresholds are configurable per project in the dashboard.
Why use an incoming webhook instead of a Slack app?
Control and simplicity. An incoming webhook is scoped to one channel you choose, revocable by you in Slack at any time, and requires granting no workspace permissions to a third party. For a signal that is just formatted messages into a channel, an OAuth app would add attack surface without adding capability.
Can I test the alert without waiting for a real incident?
Yes, the alert settings include a test button that sends a sample alert through your configured channels immediately, so you verify the webhook and email delivery in seconds rather than discovering a typo during a real runaway.
When should a cost alert fire versus a cap enforce?
Alerts inform, caps act, and mature setups use both: alerts at thresholds where you want a human judgment call, caps at the boundary where you want a guarantee regardless of who is awake. A velocity alert at 2am is only useful if someone reads it; a spending cap blocks or downgrades the user’s calls with nobody watching.
Keep reading
Wire the channel before the incident
The alert that saves you is the one configured before you needed it. Integration is about ten minutes, alerts are a webhook paste on top, and the alert history view is on the demo if you want to see the end state first.