Alerts

Slow or failing endpoints,
caught in a minute.

Write an alert as one sentence, check it against the last 24 hours of real traffic, and save it. Monocle evaluates it every minute and tells you once when it fires, once when it resolves.

Get started Read the docs

Every minute

Evaluated on a rolling window

24h preview

See when it would have fired

Threaded resolution

In Slack and Discord

From your AI tool

Created over MCP

Write it as a sentence, test it on real traffic

Open an endpoint and click Create alert: the route and service are already filled in. Pick a metric and a threshold, and the preview replays the alert on the recent traffic before you save it.

p50, p95, p99, or average latency, error rate, 5xx rate, or request rate
Windows from 5 to 30 minutes, and a minimum number of requests so quiet hours stay quiet
For latencies, a threshold suggested from your own traffic: the 24h p95 × 1.5

One message when it fires, one when it resolves

An alert sends nothing between the two, even if the value gets worse. Monocle records the peak of the incident instead, and links the message to the slowest traces of the endpoint.

Pending first, Firing only once the condition holds for the Must breach for duration
In Slack and Discord, the resolution replies in the thread of the firing message
No data received at all? The evaluation is skipped, so an ingestion delay never pages you

alert.firing

{
  "type": "alert.firing",
  "timestamp": "2026-10-17T10:00:05.123Z",
  "data": {
    "ruleName": "p95 latency · POST /api/checkout > 800ms",
    "metric": "p95",
    "comparator": "gt",
    "threshold": 800,
    "windowMinutes": 5,
    "serviceName": "api",
    "environmentName": "Production",
    "value": 912.4,
    "samples": 1530,
    "slowestTracesUrl": "https://monocle.sh/acme/..."
  }
}

Let your agent set them up

The Monocle MCP server can list, preview, create, edit, pause, and resume alerts. Ask your assistant to cover your busiest endpoints, then review what it made.

Alerts created over MCP start paused, so nothing fires before you check them
Each alert page charts every evaluation with the threshold, incidents, and deployments
Pause an alert during a planned migration, and resume it after

claude code

# Ask your assistant:
# "Add p95 latency and error rate alerts on our
#  10 busiest endpoints, sent to #alerts"

# It previews each alert on the last 24 hours,
# picks a threshold from your real traffic,
# and creates them paused, ready for you to review.

Works with

Plugs into your stack

Slack

Resolution in the same thread

Discord

Resolution in the same thread

Webhooks

Signed alert.firing and alert.resolved events

Email & in-app

For every member of the project

Monocle MCP

Create and tune alerts from your editor

Deployments

Shown on the alert chart

Next up

Every part of your app, already on a dashboard.

Endpoints, database queries, jobs, external APIs, cache, emails, and users each get their own dashboard the moment your app sends data. No panels to build, no queries to write.

Explore Dashboards

Read the docs

All features

Early access

See what your apps
are really doing.

Traces, logs, errors and metrics for every app you ship. Bring it all into focus.