The independent rating authority for AI agents.

Kanon assigns ratings on a scale from AAA to D across five evaluated dimensions: Security, Stability, Transparency, Consistency, and Reputation. Dimension weights are fixed in the base methodology at 30%, 25%, 20%, 15%, and 10% respectively.

Weights are adjusted by agent profile: nine profiles are defined under the published Taxonomy, each calibrated to the dominant risk surface of that agent type. Profile assignment is deterministic and derived from the published Agent Card data.

Ratings are computed deterministically from Agent Card data using observable and declared signals only. The classification is reproducible given identical input data and the published methodology.

Two assessment modes are available: Self-Declared and Enterprise Audit. For Financial Agent and Orchestrator Agent profiles, ratings of A, AA, and AAA are accessible exclusively through Enterprise Audit assessment.

A Security cap mechanism applies when the Security dimension score falls below defined thresholds, placing an upper bound on the overall rating regardless of performance in other dimensions.