L

Learn AI

Track progress · learn offline

Open

Lesson 5 of 7

AI agents & MCP

7 min read

A normal scenario always runs the same modules in the same order. But some jobs don't have a fixed order — so what if the AI could pick its own next step?

From fixed chain to agent

A scenario is a fixed chain: module 1, then 2, then 3, every single time. A Make [agent](glossary://agent) flips that. You give it a goal and a set of tools, and it decides — for each request — which tool to reach for and in what order. Ask it something simple and it uses one tool; ask something gnarly and it chains several. The steps aren't drawn in advance; the agent chooses them.

An agent isn't a longer chain — it's a decider. Same goal, different tools reached for, depending on what each request needs.

MCP is the universal socket

How does an agent get its tools? Through [MCP](glossary://mcp) — the Model Context Protocol, one shared standard for plugging tools into an AI. Instead of a custom hookup for every app, MCP is a single socket: any tool that speaks it — your calendar, a database, a search — snaps in the same way. One port, many tools, all reachable by the agent.

MCP means an agent's toolbox is open-ended: anything that speaks the standard plugs into the same port and becomes a tool it can choose.

An agent that acts — sends, books, pays — can act wrongly. Give it only the tools it needs, and keep a human check on anything that spends money or can't be undone.

The shape of it

Two requests hit your agent: one needs only the calendar, one needs the calendar *and* a search. What does the agent do?

Continue in the app

Take the whole Make for AI course — tracked

Get your personalized path, progress and streaks in the app — this lesson and every next one, in order.