AIR

AIR is a security platform designed to govern and protect AI add-ons across agent ecosystems. It addresses risks introduced by the extensions, integrations, and components that AI agents rely on — including skills, plugins, MCPs (Model Context Protocol servers), subagents, hooks, commands, and rules. The platform operates across three agent deployment types: - Endpoint Agents - Cloud Agents - SaaS Agents Core capabilities are organized into three functional areas: AI Posture Management: - Provides visibility into the full agent ecosystem - Continuously assesses risk across agent configurations - Detects misconfigurations before they result in security incidents Runtime Protection: - Enforces policies on agent behavior and permissions in real time - Controls what actions agents are permitted to take, evaluated at the individual action level rather than at the agent level Supply Chain Security (Marketplace): - Acts as a centralized, vetted source for AI extensions including skills, MCPs, plugins, and subagents - Supports both externally sourced and internally built add-ons - Provides a "secure by default" add-on marketplace (listed as coming soon) Scanning capabilities: - AI Add-on Scanner: Identifies hidden risks, excessive permissions, and malicious behavior in individual AI add-ons - Agent Scanner: Analyzes an agent's full set of add-ons, permissions, and configurations to surface security risks and clarify what the agent is capable of doing The platform covers seven categories of AI add-ons: Skills, Plugins, MCPs, Subagents, Hooks, Commands, and Rules.