Service

Caindrift

Detect meaningful changes in agents, models, tools, policies, and compliance state before drift becomes an incident. Creates baselines, observes current state, compares against baseline, detects drift, classifies severity, and preserves evidence. Statistical methods include PSI, Jensen-Shannon, KS test, chi-square, and mean/variance shift. Integrates with CAIN Evidence, Observability, and Compliance. Does NOT invent baselines -- caller supplies observed state and configuration.

Worked example

This service has no operations beyond health checks right now.