Smart Reports

Smart Reports is a natural language query interface on the Reports page. It lets you ask questions about your organization’s monitoring data in plain English and receive instant chart visualizations with interactive data tables. After viewing results, you can ask follow-up questions that trigger AI-powered analysis with cross-source data correlation.

Access Smart Reports from the Reports page. The query bar is at the top of the page. Requires a paid subscription (Professional tier or higher). Each follow-up analysis uses 1 AI query from your subscription pool; the initial chart query does not consume AI credits.

Asking a Question

Type a question into the query bar and press Enter or click Send. The system translates your question into a structured database query, executes it, and renders the results as an interactive chart. Example questions:

  • “Show CPU usage over the past 24 hours”
  • “Which server has the highest memory usage?”
  • “Show errors and critical events this week”
  • “What agents are installed?”
  • “Show network throughput over the past week”
  • “Which apps installed across the fleet aren’t on the approved list?” (joins installed-software inventory against the Approved Applications list maintained on the Integrations page)

Long Time Ranges

Raw health metrics are kept for 14 days. For a question about more than 7 days of metrics, or about trends across weeks or months, Smart Reports reads summaries instead of raw samples: 5-minute summaries of health metrics and 15-minute summaries of hardware readings. Each summary holds the average, minimum and maximum for its interval. A question of 7 days or less reads the raw samples.

The summaries exist for organizations on a paid retention tier (Extended, Annual or Maximum, see Data Retention), from the day the tier is active. Without one, Smart Reports is instructed to answer a longer question from the raw samples of the last 14 days and to say in the chart title that longer history needs paid metrics retention. This is an instruction to the AI that plans the query rather than a fixed limit, so check the dates in the results to see which period an answer covers.

Visualizations

Results are automatically rendered as the most appropriate chart type based on the data: line charts for time-series trends, bar charts for comparisons, doughnut charts for distributions, and stat cards for single values. You can switch between Line, Bar, Donut, and Radar chart types using the toggle buttons above the chart. A sortable data table is available below each chart.

Follow-Up Analysis

After any chart renders, an “Ask a follow-up question” input appears below the results. Type a deeper question about the current visualization, such as “What was happening on that server when CPU spiked?” The system then:

  1. Extracts context from the current chart (data source, time range, primary server, metric values)
  2. Runs 2–5 correlation queries across multiple data sources (health metrics during event windows, event activity during metric spikes, top event providers, fleet-wide comparisons)
  3. Sends all correlated data to Claude with a professional analysis prompt
  4. Streams the narrative analysis token-by-token with live markdown rendering

The analysis follows a structured format: Summary, Detailed Analysis, Root Cause (if identifiable), and Recommendations. Each analysis is persisted and included in PDF exports.

Exports

Three export options are available via buttons above the chart: PNG exports the chart image, CSV exports the raw data table, and PDF opens a print-ready report with the chart, data table, and any follow-up analysis narrative.

Contextual Follow-Up Suggestions

After each query, the system generates relevant follow-up question chips based on the data pattern. These are clickable shortcuts that run a new query immediately. For example, if your chart is grouped by server, suggestions might include “Show all metrics for [highest server]” or “What was happening on [server] when CPU spiked?”