Identity & Authorization

Why This Matters

An agent that can call tools is a principal. If it borrows a user's standing token, you have a confused deputy. Identity and authorization are not a security add-on; they are the control plane for action.

Architecture Position

The Trust Layer. Separate from general AI Security: this domain is about who the actor is and what it is allowed to do, continuously. See it on the architecture map.

Start Here

Building Enterprise AI Agents

Production harness: runtime, orchestration, tools, memory, RAG, security, observability, and production controls.

Advanced · 40 min

Core Concepts

Identity

Human identity
Agent identity
Workload identity
Service identity

Authentication

OIDC
OAuth
Service authentication
mTLS

Delegation

OAuth delegation
Token exchange
Delegation chains
Impersonation
Short-lived credentials

Authorization

RBAC
ABAC
ReBAC
Fine-grained authorization
Capability authorization

Agent Authorization

Tool permissions
Scoped execution
Continuous authorization
Step-up authorization

Enterprise Identity

Federation
Zero Trust
Policy-as-code
Access governance

Relevant Standards

OAuth
OIDC
JWT
DPoP
FAPI
Token Exchange
SPIFFE/SPIRE

Advanced

Advanced material

Deeper guides for this domain are on the way.

Coming Soon

Architecture Patterns

Applied patterns for this domain live in the pattern library.

Browse architecture patterns

Case Studies

Enterprise walkthroughs that apply this topic in a full system.

Browse case studies

Roadmaps

This topic appears in the Enterprise AI Architect path.

Follow the Enterprise AI Architect roadmap