Anomaly detection
Detectors, findings, and the methods catalog — spotting cost and behaviour changes before the invoice does.
Anomaly detection watches your telemetry and cost streams for changes that deserve attention — a spend spike, a token-usage shift, an agent behaving differently. This page outlines the section; it will be expanded as the module evolves.
What it shows
The detectors active for your organization, a feed of findings each detector has raised (what changed, when, by how much, on which service or feature), and a methods catalog documenting how each detection method works and when it fires.
Questions it answers
Did cost or usage spike outside its normal pattern? Is an agent suddenly making more calls, using a different model mix, or failing more often? Which detection methods are watching which signals?
Main interactions
Review and triage findings, tune or enable detectors, and read a method's catalog entry to understand exactly why a finding fired.