search
Ctrlk

Agentic Call Stats

The Agent Component Usage Stats tab provides a detailed view of how an agent component or MCP endpoint is being used over time. It helps identify usage patterns, detect anomalies, and understand agent behavior.

hashtag
How to Access

  1. Go to the Agentic Discovery section in the Akto dashboard

  2. Click on a Collection to view the grouped agent components

  3. Select a Component from the list

  4. In the right-hand Component details pane, navigate to the "Usage Stats" tab

hashtag
Time Range Selection

At the top of the tab, you can choose the time interval to analyze traffic trends. Available intervals include:

  • 15 minutes

  • 30 minutes

  • 1 hour

  • 3 hours

  • 6 hours

  • 12 hours

  • 1 day

  • 7 days

This allows you to view recent activity spikes or analyze long-term patterns.

hashtag
Graph 1: Component Invocation Count Over Time

What it shows: This time series chart displays the number of times the selected agent component was invoked during the chosen interval.

How to use it:

  • Spot spikes or drops in usage that may indicate incidents or load changes

  • Identify peak usage periods and correlate with deployments, issues, or user activity

  • Monitor agent component health and availability

  • Detect unusual invocation patterns that might indicate security issues

hashtag
Graph 2: Component Invocation Frequency Distribution

What it shows: This histogram visualizes how many users or agents are making a specific number of component invocations. It shows the distribution of call volume across users and systems.

How to use it:

  • Identify high-frequency users (potentially automated systems or abuse)

  • Understand general usage patterns — whether most invocations are casual or intensive

  • Pinpoint outliers and investigate unusual behavior

  • Detect potential denial-of-service or resource exhaustion attacks on agent components

hashtag
Use Cases

  • Performance Monitoring: Track agent component response times and availability

  • Security Analysis: Detect unusual invocation patterns that might indicate attacks

  • Capacity Planning: Understand peak usage to scale infrastructure appropriately

  • Anomaly Detection: Identify sudden changes in component usage patterns

Last updated