How Agent Identity Security Works

Learn how Agent Identity Security continuously discovers, assesses, governs, and controls AI agents across the enterprise.

Agent Identity Security operates as a continuous security lifecycle, not a one-time discovery sweep. Information from supported enterprise environments is normalized into the Agent Registry and AIBOM, enriched with identity and resource context, assessed for posture and risk, governed through policies and workflows, and monitored over time.

  1. Discover: Connect to supported AI platforms, developer environments, custom frameworks, and endpoints to find agents wherever they operate.
  2. Build identity and resource context: Normalize discovered agents into the Agent Registry and AIBOM. Link owners, users, credentials, models, skills, MCP servers, and connections to each agent.
  3. Assess posture and risk: Evaluate supported posture conditions, detect configuration changes, and surface risk signals.
  4. Govern: Use agent posture data, risk configuration, compliance findings, event logs, and workflows to assess and respond to issues.
  5. Secure at runtime: Apply agent policies across supported MCP, file, package, web, and terminal actions. Use the MCP Gateway for controlled MCP access.
  6. Monitor and audit: Capture activity, configuration changes, policy decisions, and enforcement outcomes to support investigation and audit.
Tip: Agent Identity Security operating model: Discover what exists → Understand identity and relationships → Assess posture and risk → Govern what requires attention → Control what agents can do → Monitor and audit continuously.