AIMentor — AI Expertise Learning Platform

Untimed. Answers are graded when you submit, with explanations and spaced-repetition scheduling.

Question 1easy
What core problem does MCP solve?
It makes LLMs generate fasterIt turns M×N bespoke tool integrations into M+N reusable connectors via a shared protocolIt replaces embeddingsIt fine-tunes models on tool usage
Question 2easy
In MCP's architecture, what is a 'server'?
The LLM itselfA separate process exposing capabilities (tools/resources/prompts) for a specific systemThe user's browserA vector database
Question 3medium
Which are MCP's three server primitives?
Embeddings, indexes, retrieversTools, resources, and promptsNodes, edges, and stateChains, agents, and memories
Question 4medium
What is the key difference between an MCP tool and an MCP resource?
Tools are faster than resourcesA tool is a model-invoked action (with side effects); a resource is app-loaded, read-only context addressed by a URIResources can only return imagesThere is no difference
Question 5medium
What message format and transports does MCP use?
GraphQL over WebSockets onlyJSON-RPC over stdio (local subprocess) or HTTP (remote)gRPC over TCP onlyPlain text over email
Question 6medium
Using FastMCP, how does a function become a tool with a schema and description?
You write the JSON schema by hand each timeYou decorate it with @mcp.tool(); the SDK infers the schema from the signature and uses the docstring as the descriptionYou train a model on the functionYou store it in a vector database
Question 7hard
What is the client's handshake sequence when connecting to an MCP server?
initialize → list (tools/resources) → callembed → retrieve → generatecompile → run → checkpointtrain → evaluate → deploy
Question 8medium
Does the host's LLM talk to MCP servers directly?
Yes, the model opens sockets to serversNo — the client inside the host mediates every call between the model and the serverOnly for resourcesOnly over HTTP
Question 9hard
When should you build an MCP server instead of a plain in-process function tool? (Select all that apply)
The capability should be reused across multiple apps/hostsIt needs to run as a separate process (different language, isolation, independent deployment)You want to plug into the MCP ecosystem (Claude Desktop, IDEs, other agents)It's a trivial helper used by exactly one app once
Select all that apply.
Question 10hard
A Python Postgres tool must also be callable from a Node.js chatbot. Why does MCP fit?
MCP compiles Python to JavaScriptA separate-process protocol server lets a Node client reuse the Python capability without reimplementing itNode cannot use databases otherwiseIt fine-tunes a cross-language model
Question 11medium
Which control model matches each primitive?
Tools = model-driven, Resources = app-driven, Prompts = user-drivenTools = user-driven, Resources = model-driven, Prompts = app-drivenAll three are model-drivenAll three are user-driven
Question 12medium
MCP turns an M×N integration problem into what (write it as a sum, e.g. 'X+Y')?
Question 13hard
Which MCP primitive (one word) represents read-only data the host loads into the model's context, as opposed to a model-invoked action?

0 / 13 answered

MCP — Model Context Protocol — Practice quiz