Observability Dashboard aggregates runtime monitoring data for agents in AgentCore. The dashboard helps you review request quality and resource consumption for tokens, models, and tool calls. On a single page, you can view Request Overview, Token Analysis, Model Analysis, and Tool Call Analysis by time range.
The UI labels and metric definitions in this topic are subject to the actual console interface.
Prerequisites
You are logged on to the AgentCore console.
View monitoring data
In the left-side navigation pane, choose Observability Dashboard to open the AgentCore Runtime Monitoring page. The page displays four monitoring modules: Request Overview, Token Analysis, Model Analysis, and Tool Call Analysis.
In the Time Range selector in the upper-right corner of the page, select the time range to include in the statistics.
To retrieve the latest displayed data immediately, click Refresh.
Review the monitoring modules, metric cards, and trend charts on the page to understand the runtime conditions during the selected time range. If no call data exists within the selected time range, the metrics and charts display No Data.
Observability Dashboard data may have a minutes-level aggregation and display delay. When troubleshooting real-time issues, consider how this delay affect the data displayed on the current page.
Understand the monitoring information on the page
Trend charts with / 2m in the title display data aggregated at two-minute intervals. Each metric card provides an entry point to its description: click the question mark icon to view how the corresponding metric is defined.
Request Overview
Metric | Description |
Total Requests | The total number of requests within the selected time range. |
Error Rate | The percentage of failed requests within the selected time range. |
Requests and Errors | The trend of request and error counts over time. |
Average Request Duration | The trend of the average duration of a single request over time. |
If Total Requests or Error Rate is abnormal, check the trends in Requests and Errors and Average Request Duration. These trends help you identify when the anomaly occurred and whether request duration changed at the same time.
Token Analysis
Metric | Description |
Token Usage Distribution by Model | The percentage distribution of token usage across models. |
Token Usage Trend by Model | The trend of token usage for each model over time. |
Use Token Analysis to compare token usage across models and identify when usage changes occur.
Model Analysis
Metric | Description |
Model Call Distribution | The percentage distribution of calls across models. |
Model Call Trend | The trend of calls for each model over time. |
Use Model Analysis to identify models with a high share of calls.
Tool Call Analysis
Metric | Description |
Tool Call Distribution | The percentage distribution of calls across tools. |
Tool Call Trend | The trend of calls for each tool over time. |