AI Hints
Agentic AI · Lesson 96 of 100

Quiz: Graphs and Teams

LangGraph, multi-agent, MCP.

⏱ 2 min read▶ Video at 2:55:54🎓 Free
Watch this lesson in the Agentic AI courseWatch on YouTube ↗Full Agentic AI course

This round covers LangGraph, teams of agents, and MCP. These come up often in interviews.

Quiz: Graphs and Teams
From the video: Quiz: Graphs and Teams
  • LangGraph
  • Teams of agents
  • MCP

Question 1

Question 1. What are LangGraph's three core ideas? A, Prompts, tokens, models. B, State, nodes, edges. C, Tasks, crews, tools. D, Files, folders, links. Pause the video and choose. The answer is B. Everything else builds on those.

Quick check
What are LangGraph's three core ideas?

Question 2

Question 2. A node returns a value for a field with no reducer. What happens? A, It's added. B, It replaces the old value. C, It errors. D, It's ignored. Pause the video and choose. The answer is B. Reducers change that.

Quick check
A node returns a value for a field with no reducer. What happens?

Question 3

Question 3. What does interrupt need in order to work? A, A checkpointer. B, A GPU. C, A web server. D, Streaming. Pause the video and choose. The answer is A. State must be saved.

Quick check
What does interrupt need in order to work?

Question 4

Question 4. On resume, where does a node restart? A, At the interrupt line. B, From its first line. C, At the next node. D, At the graph start. Pause the video and choose. The answer is B. So keep side effects after it.

Quick check
On resume, where does a node restart?

Question 5

Question 5. What is a thread ID for? A, Parallel speed. B, Separating saved conversations. C, Billing. D, Logging only. Pause the video and choose. The answer is B. Like a save file.

Quick check
What is a thread ID for?

Question 6

Question 6. What's the main risk of too many approval requests? A, High cost. B, People stop reading them. C, Slow models. D, Bad prompts. Pause the video and choose. The answer is B. That's approval fatigue.

Quick check
What's the main risk of too many approval requests?

Question 7

Question 7. When is a supervisor better than handoffs? A, One specialist is enough. B, Results from several must combine. C, There are no tools. D, Speed is everything. Pause the video and choose. The answer is B. It stays in charge.

Quick check
When is a supervisor better than handoffs?

Question 8

Question 8. What should agents share with each other? A, Everything. B, Conclusions, not working notes. C, Nothing. D, Only passwords. Pause the video and choose. The answer is B. Clean context.

Quick check
What should agents share with each other?

Question 9

Question 9. MCP connects an agent to what? A, Other agents only. B, Tools and data. C, GPUs. D, Browsers only. Pause the video and choose. The answer is B. A two A connects agents.

Quick check
MCP connects an agent to what?

Question 10

Question 10. In MCP, who controls when a resource is read? A, The model. B, The application. C, The server. D, Nobody. Pause the video and choose. The answer is B. Tools are model controlled, prompts are user controlled.

Quick check
In MCP, who controls when a resource is read?

Summary

Seven or more is very strong. Now the final round.

Key takeaways
  • 7 or more: very strong
  • Reducers and interrupts matter
  • Next: safety and evals
Next up: Quiz: Safety and Evals — the professional round