Skip to content
NTPU AI Learn

AI 201 Practice

Ask, Inspect, then Connect: from using MCP to understanding it

Info

Sandbox (illustrative data). This is a SANDBOX. Every record below is illustrative sample data, not live NTPU records. The real service is NTPU AI4X MCP; Practice C shows you how to reach it.

Practice A: Ask

Experience what an AI answer looks like when it is backed by an MCP Tool. Pick a sandbox question and run it.

  1. Choose one of the sandbox questions and run it.
  2. Read the final answer as if you were the person who asked.
  3. Ask yourself: can I tell where this information came from, and how fresh it is?
  4. Write down what you cannot tell yet. Practice B will help you find it.
Pick a question

Practice B: Inspect

Move from user to someone who understands how the tool works. Look only at what the interface actually shows: the tool selection, the arguments, the tool result and its provenance. Do not guess at what the model "thought". Try each prompt yourself before opening the guidance.

Tip

Ask a question above first, then work through the checklist.
  1. Which Tool was used?
  2. What is the Tool description?
  3. What are the arguments?
  4. Which fields does the result contain?
  5. What is the Source?
  6. Is there provenance?
  7. Which content is Fact?
  8. Which content is AI synthesis?

Practice C: Connect

Now connect a real MCP Client to NTPU AI4X MCP. Learn teaches the concepts; AI4X MCP runs the service. The step-by-step setup guide lives in the canonical technical documentation so it can stay current.

  1. Choose an MCP-capable Client that you are allowed to use.
  2. Follow the current NTPU AI4X MCP connection guide for that Client.
  3. In the Client, run discovery (the tools/list equivalent) and read the tool names and descriptions.
  4. Call one tool with valid arguments from its input schema.
  5. Verify provenance: source, retrieval time, last-updated time and canonical URL. Then decide what you would still check by hand.
  • ChatGPT: Check the Client’s current MCP support. Its UI changes quickly, so there are no screenshots here.
  • Claude: Check the Client’s current MCP support. Its UI changes quickly, so there are no screenshots here.
  • VS Code: Check the Client’s current MCP support. Its UI changes quickly, so there are no screenshots here.
  • Claude Code: Check the Client’s current MCP support. Its UI changes quickly, so there are no screenshots here.
  • Other MCP-capable Client: Any Client that supports MCP can work. Check its current MCP support first.

Info

This is a simulated rehearsal. It never reaches a real server.

Rehearse a real connection in order: review consent first, then discover tools, build arguments, and finally verify the result. Nothing here reaches a real server and no credential is needed. After the rehearsal, connect your own MCP client for real by following the current NTPU AI4X MCP guide.

Step 1: Review consent and safety

Step 1 of 4 Review consent and safety

Before connecting, confirm what you are willing to allow. Tick every item to continue.

Step 2: Discover tools and choose

Step 2 of 4 Discover tools and choose

Complete the previous step first.

Step 3: Build arguments and call

Step 3 of 4 Build arguments and call

Complete the previous step first.

Step 4: Verify the result

Step 4 of 4 Verify the result

Complete the previous step first.

Info

Learn teaches, AI4X MCP runs. The setup guide lives in the canonical technical documentation, which is always the latest version.

Open the NTPU AI4X MCP connection guide (opens in a new tab)

Last verified: 2026-09-19. Tested with: MCP specification: see AI4X MCP docs / Client: see docs / Server: NTPU AI4X MCP

Sync my progressJoin a session to keep progress across devices and share it with your instructor

Info

After you join, your completed lesson numbers are sent to your instructor together with your learner code, so they can see your progress. Use the codes your instructor gave you, and do not enter your name, company secrets or other personal data. Progress is kept with the rest of that session's data.

What next?