Core Concepts

Learn the essential terms and architecture concepts for Agent Identity Security.

Before you work with Agent Identity Security, it helps to understand the key terms and the mental model behind them. These concepts appear throughout the product and its documentation.

Agents and Identity


Concept	What it means in Agent Identity Security
AI agent	An AI-powered system that uses models, instructions, and tools to perform tasks or take actions on behalf of a user or a business process. In Agent Identity Security, an agent is also an enterprise identity because it can use credentials, receive permissions, and interact with enterprise resources.
Agent identity	The identity context associated with an agent: which agent is acting, on whose behalf, using which identity, and with access to which resources.
Agent owner and user	The owner holds accountability for an agent. Associated users provide context about the people who interact with or relate to that agent. Enterprise identity integrations can enrich these relationships with organizational and lifecycle context.
Agent Registry	The centralized inventory of agents that Agent Identity Security has discovered, including identity, ownership, platform, posture, access, connections, and activity context.
AI Bill of Materials (AIBOM)	The broader inventory and relationship context for the agent ecosystem: agents, MCP servers, skills, models, users, credentials, and connections.
Credentials	Authentication material associated with agents or their interactions—such as tokens or keys. Credential relationships help establish identity and access context for an agent.
Models	The AI models that agents use. Agent Identity Security provides inventory and relationship context so you can see which models are associated with which agents and apply supported governance decisions.
Skills	Capabilities or instructions that extend what an agent can do. Agent Identity Security provides visibility into discovered skills and their relationship to agents, with governance actions available where supported.

Note: The Agent Registry answers What agents exist? The AIBOM answers What is associated with each agent, and how is it connected?

Security and Control


Concept	What it means in Agent Identity Security
MCP servers and tools	Model Context Protocol (MCP) lets AI applications and agents interact with external tools and resources through a standardized protocol. Agent Identity Security can inventory MCP servers and their relationships with agents.
Agent posture	The security and configuration state of an agent, based on Agent Identity Security posture checks across areas such as ownership, model and prompt drift, authorization changes, authentication configuration, and MCP security.
Agent risk	A prioritization mechanism based on security context. Agent Identity Security evaluates risk signals across identity, data, posture, access, and user dimensions, with configurable weighting.
Agent policy	A runtime security policy that defines what supported agents are permitted to do. Agent Identity Security supports Allow, Deny, and Require Approval decisions across supported MCP, file, package, web, and terminal actions.
MCP Gateway	An inline security control point between supported MCP clients and MCP servers. It provides authentication, policy enforcement, activity capture, and supported response-protection controls.
Human-in-the-loop (HITL)	A control mechanism that pauses a selected agent action and routes it to a configured approval channel before it can proceed.
Workflows	Automated or approval-driven processes that connect agent, identity, posture, and lifecycle events with governance actions such as ownership transfer, agent disablement, token revocation, and notifications.