Airia

AARM Core

Enforce AARM at a single inline interception point — the Unified AI Gateway

Visit website ↗

Classification

Coverage surface: MCPAPI
Stage: Launched
Type: Commercial
Target audience: Enterprise
Deployment: SaaSSelf-hosted

Technical profile

Spec-grounded axes, verified by the TWG.

Interception architecture (R1): Protocol Gateway
Policy model (R3): Hybrid
Authorization decisions (R4): ALLOWDENYMODIFYSTEP_UPDEFER

Conformance review

`R1`	Pre-execution interception	Pass
`R2`	Context accumulation	Pass
`R3`	Policy evaluation with intent alignment	Pass
`R4`	Five authorization decisions	Pass
`R5`	Tamper-evident receipts	Pass
`R6`	Identity binding	Pass
`R7`	Semantic distance tracking	Not met
`R8`	Telemetry export	Not met
`R9`	Least privilege enforcement	Not met

Platform capabilities

Single inline interception point (Unified AI Gateway) between the calling agent and every downstream model, tool, or provider
No action reaches a tool endpoint without first passing policy evaluation
Per-session context accumulation checked on every intercepted action
Inline DLP inspection layer classifies each intercepted action
Five-outcome Policy Engine: allow, deny, modify, step-up, defer
Deferrals and step-ups suspend on a bounded timeout that resolves to deny
Fails closed on any incomplete or failed evaluation

Architecture

Airia enforces AARM at a single inline interception point: the Unified AI Gateway, which sits between the calling agent and every downstream model, tool, or provider. No action reaches a tool endpoint without first passing policy evaluation. Each intercepted action is checked against accumulated per-session context, classified by an inline DLP inspection layer, and evaluated by a Policy Engine that returns one of five authorization decisions: allow, deny, modify, step-up, or defer. Deferrals and step-ups suspend on a bounded timeout that resolves to deny, and any incomplete or failed evaluation fails closed rather than releasing the action.