18 MCP Topics, Each Filed Under a Task Class
The Model Context Protocol is an open specification for connecting an agent application to capabilities outside itself. It fixes three roles — a host application, a client holding one connection, and a server exposing capabilities — an opening handshake in which each side declares what it supports, and a short list of things a server may publish: operations to invoke, content to read, and message templates to fill in. Everything else is left to whoever builds the server.
Each page below explains one term in that vocabulary at the protocol level, and then states the task class this registry assigns to work involving it. The explanations are the point of the pages and are written to be correct; nothing on them describes what any particular product supports, because a description of a product expires and a description of a role does not.
The Registry makes no representation that an assessment predicts the outcome of any task, and no assessment is validated against real-world results.
The vocabulary, in the order the pieces meet
Read top to bottom this is one sentence: a host runs a client, the client connects to a server, the server publishes tools, resources and prompts, the connection runs over a transport, and a catalogue is how the host learned the server existed.
- Host
- The application a person actually uses. It runs the clients, holds the conversation with a language model, and decides which connected capabilities are put in front of it.
- Client
- One connection to one server, held inside the host. It performs the opening handshake, sends requests, and receives responses and notifications for the life of the session.
- Server
- The process on the other end of that connection. It advertises what it offers, answers requests against some underlying system, and holds whatever credentials reaching that system requires.
- Tool
- A named operation a server offers for invocation, published with a description and a schema for its arguments so that a caller which has never seen it before can still call it correctly.
- Resource
- Content a server makes readable, identified by a URI. A resource is addressed and read; a tool is invoked. The distinction is the reason both exist.
- Prompt
- A named, reusable message template a server publishes, filled in from arguments and returned ready to send. Typically chosen by a person rather than selected automatically.
- Transport
- The channel the messages travel over, kept separate from the messages themselves — which is why the same operations survive a move from a local process to a network connection.
- Registry
- A catalogue of servers and their connection details. It settles which server and how to reach it, before any connection is opened; it settles nothing about whether the entry is accurate.
All 18 topics
Every row names a topic, the task class the Registry assigns to work involving it, and a link to the page that explains it. The task class is what the order form consumes — the link on each topic page selects it there already.
| Topic | Task class | Page |
|---|---|---|
| What Is an MCP Server? | Code modification | Read the definition → |
| What Is an MCP Host? | Code modification | Read the definition → |
| What Is the Model Context Protocol? | Information retrieval | Read the definition → |
| What Does Inspecting an MCP Server Involve? | Data processing | Read the definition → |
| What Are Tools in MCP? | Code modification | Read the definition → |
| How Does MCP Differ From an API? | Data processing | Read the definition → |
| What Is an MCP Client? | Code modification | Read the definition → |
| What Is an MCP Registry? | Information retrieval | Read the definition → |
| How Are MCP Servers Listed and Discovered? | Information retrieval | Read the definition → |
| How Does Authentication Work in MCP? | Code modification | Read the definition → |
| What Are Resources in MCP? | Information retrieval | Read the definition → |
| What Are the Security Considerations in MCP? | Data processing | Read the definition → |
| What Is a Remote MCP Server? | Code modification | Read the definition → |
| What Does Building an MCP Server Involve? | Code modification | Read the definition → |
| What Does Implementing MCP in Python Involve? | Code modification | Read the definition → |
| What Are Prompts in MCP? | Code modification | Read the definition → |
| What Is a Transport in MCP? | Code modification | Read the definition → |
| What Does Implementing MCP in TypeScript Involve? | Code modification | Read the definition → |
Where this vocabulary lands on the order form
The engine behind every assessment on this site recognises a fixed set of task classes, and each topic above is assigned to exactly one of them on the reasoning stated on that topic's own page. The table below counts how the assignment fell out. It is concentrated rather than even, and deliberately so: work on a protocol integration is engineering work, and a class chosen to make a table look balanced would be an invented fact about the task.
| Task class | Covers | Topics assigned here | Task-class page |
|---|---|---|---|
| Code modification | Writing, editing, refactoring, and testing code | 11 of the 18 | View task class → |
| Information retrieval | Search, information gathering, research, summarisation | 4 of the 18 | View task class → |
| Data processing | Data handling, transformation, analysis, and computation | 3 of the 18 | View task class → |
The classes absent from this table are not absent from the order form. Every task class the engine recognises is selectable there, whether or not any page in this family happens to use it.
Frequently asked questions
- What is the Model Context Protocol?
- It is an open specification for connecting an agent application to capabilities that live outside it. The specification defines three roles — a host application, a client that holds one connection, and a server that exposes capabilities — an opening handshake in which each side declares what it supports, and a small set of things a server may publish: tools to invoke, resources to read, and prompt templates to fill in. It does not define what any server exposes, what language it is written in, or which capability an application should use.
- Does the Registry publish a verdict for the topics listed here?
- No. These pages define the vocabulary and state how the Registry would classify a task involving it. The Registry makes no representation that an assessment predicts the outcome of any task, and no assessment is validated against real-world results. A verdict, a recommended execution window and the named risk factors are the content of a numbered assessment ordered against a buyer’s own task description.
- How does an MCP topic become a task class on the order form?
- Every topic page states the task class the Registry assigns to work of that kind and links to the order form with that class already selected. The assignment is fixed per topic, published on the topic’s own page, and reproduced in full in the table on this page. Most of this vocabulary describes engineering work, which is why the assignment is concentrated rather than spread evenly across the classes.
Get an assessment
The pages in this family are free and explain a protocol. What the Registry sells is something else: a permanent, numbered task risk assessment for one submitted task at one model configuration, computed by the deterministic method published in full at /method — a verdict on a seven-point scale, a recommended execution window, and, above the first tier, the named risk factors the class tracks. EUR 1.90 Standard, EUR 4.90 Extended, EUR 14.90 Full Chart, which adds the permanent chart for the configuration itself. Assessments from EUR 1.90; machine-readable at /pricing.json.