IntegrAuth Academy
Identity, security and practical AI — free, from zero. Byte-sized lessons, real diagrams, hands-on labs, zero fluff. Two programs, each with its own exam and certificate.
English is hard? Your browser can translate this page into Tamil, Telugu, Kannada, Malayalam or Hindi. Tech words stay in English so you learn the real names.
How?
- Chrome on phone: ⋮ menu → Translate
- Chrome on computer: right-click → Translate to …
- Edge: the translate icon in the address bar
Start here: AI agents that act
Meet Kai, an AI agent. Kai reads a request, makes a plan and uses real tools to do the work, often for a person. Bot A, our RPA bot, only follows a fixed script. Your goal: learn how to give an AI agent an identity, limits and an off switch.
Why this track matters
An agent that can read an invoice can also leak one. So we control the tools Kai uses, the data Kai reads and the actions Kai takes.
Picture Kai as a new assistant
Imagine you hired a helper today. You would not hand over your house keys and bank PIN on day one. You would do four things.
📜 Give a letter of authority
It says what the helper may do, for whom and for how long. Kai gets an identity and limits.
✋ Say "ask me before you spend"
For money, the helper stops and asks you first. Kai waits for a person to approve.
📒 Ask for a diary
The helper writes down every step. Nobody can quietly erase a page. Kai leaves an audit trail.
🛑 Keep a way to say stop
If the helper goes wrong, you cancel the letter at once. Kai has a kill switch.
New to agent or MCP? Foundations explains them. New to AI? Try AI from zero first. It is optional.
The journey
- Governing MCP — one guarded door for tool calls.
- Fine-grained authorization — can Kai open this record?
- Permission-aware RAG — AI reads only what the person may.
- Human-in-the-loop — a person approves risky actions.
- Agent registry & kill switch — list agents and stop one fast.
- Delegated access — use other accounts without passwords.
- A secure copilot, step by step — all controls in one flow.
- Prompt injection — text can give the agent new orders.
- Poisoned memory and tools — notes and tools that change later.
- Agent-to-agent — keep consent across agent hops.
- Agent audit trails — prove what an agent did.
- Cheat sheet & pop quiz — a checklist and five scenarios.
How to use it
Take one lesson at a time. Progress saves in your browser. Most lessons end with a lab.
You will give an agent its own identity. You will keep it to least privilege on tools and data. You will keep a human on risky actions. You will switch off a bad agent fast. These are the controls behind a copilot you can trust.
Start with governing MCP.