3.11 — Agents Quiz 📊

1. What’s the difference between a “tool” and “MCP”?

2. Name the 4 steps of the ReAct loop, in order.

3. You want a team of agents organized by roles (manager, researcher, writer), simply. Which framework?

4. Which framework keeps a state memory and resumes after a crash?

5. True or false: the more agents you add, the better the result.

6. Give an example of a task where a single agent is plenty.

7. You want to automate visually, without coding, with a maximum of turnkey integrations. Make or n8n?

8. In the case study (3.9), which tool serves as memory / database for the agent?

9. True or false: n8n works only in the cloud.

👉 Show answers

1. The tool is the ability to call an external function; MCP is the universal standard for plugging in those tools/sources without a custom integration.
2. Reason → Act → Observe → Loop.
3. CrewAI (roles, the most intuitive).
4. LangGraph (checkpointing / state memory).
5. False. Multi-agents cost much more and often degrade sequential tasks. Start simple.
6. A calendar assistant, a one-off web search, a document summary… any well-defined, low-tool-use task.
7. Make (visual, mainstream, 3000+ integrations). n8n is more technical/powerful.
8. Airtable (the base / memory where the agent reads and writes).
9. False. n8n is used in cloud OR self-hosted (on your premises, for privacy).