​
Alerts

AgentMark’s alert system allows you to monitor critical metrics for your applications and get notified when certain thresholds are exceeded. This helps you stay informed about your application’s performance and potential issues. Alerts Dashboard

​
Overview

Alerts in AgentMark help you:
  • Monitor important metrics like cost, latency, and error rates
  • Set thresholds for acceptable values
  • Define time windows for measurement
  • Receive notifications via Slack or webhooks when thresholds are exceeded
  • Track alert history to analyze patterns

​
Available Metrics

AgentMark currently supports alerts on the following metrics:
MetricDescription
CostMonitor the total cost of your API calls to LLM providers
LatencyTrack response times for your AI requests
Error RateMonitor the percentage of failed requests

​
Creating an Alert

To create a new alert:
  1. Navigate to your application’s dashboard
  2. Click on the “Alerts” tab in the navigation menu
  3. Click the “Create Alert” button
  4. Configure the alert with the following settings:
    • Name: A descriptive name for the alert
    • Metric: Select the metric to monitor (cost, latency, or error rate)
    • Threshold: Set the threshold value that will trigger the alert
    • Time Window: Define the period over which the metric is measured

​
Notification Options

AgentMark offers multiple ways to receive alert notifications:

​
Slack Integration

To receive alerts via Slack:
  1. Enable the “Send to Slack” option when creating an alert
  2. If you haven’t connected Slack yet, click the “Connect Slack” button
  3. Select the Slack channel where notifications should be sent

​
Webhooks

For custom integrations, you can use webhooks:
  1. Enable the “Use Webhook” option when creating an alert
  2. Configure your webhook endpoint in the Developers section of your application
  3. Alerts will be sent as HTTP POST requests to your endpoint with alert details in the request body
For detailed information on implementing and configuring webhooks, including the exact payload format and handling instructions, see the Test Webhook documentation.

​
Alert Status

Alerts have two possible states:
  • Triggered: When the monitored metric exceeds the defined threshold
  • Resolved: When the metric returns below the threshold

​
Alert History

For each alert, you can view its history to analyze patterns and understand when and why alerts were triggered:
  1. Navigate to the “Alerts” tab in your application dashboard
  2. View the list of alerts
  3. Click on an alert to see its history, including:
    • When the alert was triggered
    • The actual value that triggered the alert
    • When the alert was resolved

​
Best Practices

For effective alert management:
  • Set meaningful thresholds based on your application’s normal behavior
  • Use appropriate time windows to avoid alert fatigue from temporary spikes
  • Configure integrations with your team’s communication tools
  • Regularly review alert history to adjust thresholds as your usage patterns change

​
Next Steps