Skip to content
AdminKit

Prompt (MCP)

What is Prompt (MCP)?

A prompt, in the Model Context Protocol, is a predefined, parameterized template a server exposes for a client to surface to its user -- as a slash command, say -- distinct from a model-invoked tool or a data resource. AdminKit's MCP endpoint check surfaces a server's exposed primitives for review, prompts included.

By AdminKit team, Netbase JSC · Updated

A prompt in the Model Context Protocol — the protocol Anthropic introduced to standardize how applications connect context and tools to an LLM — is a predefined, parameterized template a server exposes for a client to surface to its user — for example as a slash command or menu item — to kick off a specific interaction, distinct from a tool, which a model invokes on its own, and a resource, which supplies reference data. A client lists a server’s available prompts, then retrieves a chosen one by name along with whatever arguments it accepts, and the server returns the actual message content to send to the model; a server can also notify a client when its prompt list changes. Each of those steps is an ordinary JSON-RPC 2.0 message: prompts/list and prompts/get are requests, and the change signal is a notifications/prompts/list_changed notification. A client can autocomplete those arguments using the protocol’s completion request, the same one a resource template’s URI variables use. Because prompts are user-controlled rather than model-controlled, a server with a seemingly reasonable set of exposed prompts still deserves the same review as its tools, since a prompt’s underlying template can embed the same kind of unreviewed instructions a malicious tool description would carry.

AdminKit groups prompt (MCP) under its AI agents & MCP cluster — browse the AdminKit glossary index for other related terms.

Frequently asked questions

How does an MCP prompt differ from a tool?

A prompt is user-controlled, surfaced to the user as something like a slash command, while a tool is model-invoked -- the model decides to call it, not the user.

How does a client discover and use a server's prompts?

It lists the server's available prompts, then retrieves a chosen one by name along with whatever arguments it accepts, and the server returns the actual message content to send to the model.

Why do a server's prompts deserve the same review as its tools?

A prompt's underlying template can embed the same kind of unreviewed instructions a malicious tool description would carry, even though prompts look user-controlled.