Alerts

Alerts watch your calls and tell you when something needs attention: more calls failing than usual, a tool your agent depends on returning errors, callers getting frustrated, or call volume dropping to nothing. Each alert sends one message when it fires and one when it resolves.

Create an alert

Dashboard: open Alerts in the sidebar and click New alert. Pick what to watch, which agent’s calls (or all agents), the threshold and the time window, and where to send it. API:

curl -X POST https://api.revring.ai/v1/alert-rules \
  -H "x-api-key: YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "Failed calls on the booking line",
    "agentId": "AGENT_ID",
    "metric": "call_failure_rate",
    "comparator": "above",
    "threshold": 20,
    "windowMinutes": 60,
    "minCalls": 10,
    "webhookUrl": "https://example.com/revring-alerts",
    "slackWebhookUrl": "https://hooks.slack.com/services/T000/B000/XXXX"
  }'

Field	Description
`name`	Shown in notifications and the dashboard.
`metric`	What to watch (see below).
`agentId`	Only this agent’s calls. Leave it out for all agents.
`comparator`	`above` (default) or `below`.
`threshold`	A percent from 0 to 100 for rates, or a number of calls for `call_count`.
`windowMinutes`	The period looked at: `15`, `60` (default), `240` or `1440`.
`minCalls`	Rates are only judged with at least this many calls in the period (default 10), so one bad call in a quiet hour does not fire an alert.
`webhookUrl`	An https endpoint for a JSON notification.
`slackWebhookUrl`	A Slack incoming webhook URL.
`enabled`	Turn the alert off without deleting it.

Each alert needs a webhook, a Slack webhook, or both. An organization can have up to 20 alerts.

What you can watch

Metric	Value
`call_failure_rate`	Percent of finished calls that failed. Outbound calls that were not answered count as failed.
`evaluation_fail_rate`	Percent of calls where at least one success criterion was judged `fail` (calls with no `pass` or `fail` verdict are not counted).
`negative_sentiment_rate`	Percent of calls with negative caller sentiment.
`tool_error_rate`	Percent of tool calls (custom tools and webhooks) that returned an error or timed out.
`webhook_failure_rate`	Percent of post-call webhooks that could not be delivered after all retries.
`call_count`	Number of calls. Use `below` to hear when calls stop coming in.

How alerts are checked

Every alert is checked every 5 minutes over its time window. When the value crosses the threshold the alert fires and sends one notification. It sends nothing more while it stays over the threshold, then sends one more when it resolves. Changing an alert’s condition takes effect at the next check: an alert that no longer meets its new condition resolves. To see an alert’s current value without sending anything, call GET /v1/alert-rules/{id}/value. To check your endpoints, POST /v1/alert-rules/{id}/test (or the send icon in the dashboard) sends a test notification. GET /v1/alert-rules/{id}/events lists past notifications with the delivery result for each channel.

Webhook payload

{
  "event": "alert.triggered",
  "alert": {
    "id": "cmuw...",
    "name": "Failed calls on the booking line",
    "metric": "call_failure_rate",
    "comparator": "above",
    "threshold": 20,
    "windowMinutes": 60,
    "agentId": "AGENT_ID"
  },
  "value": 27.5,
  "sample": 40,
  "summary": "Failed calls: 27.5% over the last 1 hour (40 in the sample); alert when above 20%.",
  "timestamp": "2026-10-06T14:05:00.000Z"
}

event is alert.triggered, alert.resolved or alert.test. sample is the number of calls (or tool calls, or webhooks) the value was computed from. Alert webhooks are signed like your other webhooks when you have a signing secret. Slack messages carry the alert name and the same summary line.

Getting started

Platform

Create an alert

What you can watch

How alerts are checked

Webhook payload

​Create an alert

​What you can watch

​How alerts are checked

​Webhook payload

Create an alert

What you can watch

How alerts are checked

Webhook payload