Module 0
✓
Setup
Build the local lab — Keycloak, Vault, OpenFGA, and the CLIs every later lesson assumes.
AskAiml Academy · Advanced
IAM for AI Agents
A standards-anchored, deeply technical course on securing AI agents: agent identity models, OAuth flows built for agents, the MCP authorization spec, scoped tool access, secrets brokering, audit and provenance, attacks and defences, governance, and the Microsoft Entra Agent ID platform.
- 11 modules
- 25 lessons
- 4h of reading
- 6 projects
0%
complete
- Lessons done
- 0 / 25
- Badges earned
- 0 / 12
- Certificate
- Not yet earned
Curriculum
Eleven modules, from primitives to production
Every module ends in a quiz. Pass it to earn the badge. Pass the final exam to earn the certificate.
Module 1
✓
Foundations
The primitives every agent-IAM bug traces back to: AuthN vs AuthZ, identity models, and a threat model for agents.
Module 2
✓
Agent identity
Give an agent a real identity — registered clients, per-agent principals, and cryptographic workload identity.
Module 3
✓
OAuth for agents
The five flows that actually matter for agents, plus sender-constrained tokens and rich authorization requests.
Module 4
✓
MCP authorization
The Model Context Protocol authorization spec, and how to ship an MCP server that survives a pentest.
Module 5
✓
Secrets & credentials
Stop handing agents raw API keys. Broker credentials just in time, scoped and revocable.
Module 6
✓
Tool authorization
Per-tool policy as code, capability tokens, and pulling a human into the loop for the dangerous calls.
Module 7
✓
Audit & observability
Prove what an agent did, why it was allowed, and reconstruct the decision months later.
Module 8
✓
Attacks & defenses
Prompt injection, token theft, confused deputy, and the MCP supply chain — with the defences that hold.
Module 9
✓
Governance & compliance
EU AI Act, NIST AI RMF, ISO 42001, and SOC 2 — translated into controls an engineer can implement.
Module 10
✓
Microsoft Entra Agent ID
The vendor deep dive: Agent 365, the four Entra agent object types, Conditional Access, ID Protection, and governance.
- Microsoft Agent 365 and Entra Agent ID landscape
- Entra Agent ID object model
- Agent OAuth flows in Microsoft Entra
- Authorization and permissions in Entra Agent ID
- Conditional Access for agents
- ID Protection for agents (Risky Agents)
- Governance and lifecycle for agent identities
- Agent 365 SDK, observability, and network controls
- Third-party agents and federation
Achievements
Collect a badge per module
Badges unlock as you pass each module quiz. The final exam issues a verifiable certificate you can share.
- Setup specialist
- Foundations specialist
- Agent identity specialist
- OAuth for agents specialist
- MCP authorization specialist
- Secrets & credentials specialist
- Tool authorization specialist
- Audit & observability specialist
- Attacks & defenses specialist
- Governance & compliance specialist
- Microsoft Entra Agent ID specialist
- Certified Agent IAM Engineer
Portfolio
Six projects that prove you can build it
Reading is not evidence. Each project is a deployable system you can put in front of an interviewer.
Agent + On-Behalf-Of + DPoP
An agent that exchanges a user token for a downstream token and binds every call to a private key.
MCP server marketplace with OAuth + signing
A registry of MCP servers with dynamic client registration, signature verification, and tool pinning.
Credential broker for tools
A broker that mints just-in-time, narrowly scoped credentials per tool call and logs every issuance.
Human-in-the-loop approval platform
Out-of-band approvals via CIBA with budgets, timeouts, and a full decision log.
End-to-end Agent IAM platform
The capstone: identity, policy, brokering, approvals, and audit in one deployable platform.
Entra Agent ID + Agent 365 governance platform
The same capstone goals implemented natively on the Microsoft stack.