An MCP prompt is a prompt template or prompt resource exposed through the Model Context Protocol so an app or agent can fetch and use a standardized prompt from a server instead of hard-coding it.
MCP is designed to let tools, models, and clients work together through a common protocol. Prompts are one part of that: they let a server publish reusable instructions, examples, or task templates that a client can retrieve dynamically.
You’d use an MCP prompt when you want:
In practice, this is useful when the “right” instructions depend on the connected system, such as a codebase, a knowledge base, or an internal workflow.
An MCP server can advertise prompt resources to an MCP client. The client can inspect what prompts are available, then request one by name, often with arguments. The server returns the prompt content, which the client can then pass to the model.
The prompt is usually not a free-form chat message. It is a structured, reusable instruction bundle: for example, a system-style instruction, a task template, or a few-shot example set. The exact shape depends on the server and the MCP implementation, but the key idea is that the prompt is discoverable and retrievable over MCP.
This is different from “prompting” in the general LLM sense. Here, “MCP prompt” means a prompt that is part of the MCP protocol surface, not just text someone typed into a chat UI.
A developer tool exposes a prompt called review_diff:
review_diffreview_diff with arguments like repository name or file pathYou are reviewing a code diff for correctness, security, and maintainability.
Focus on breaking changes, missing tests, and unsafe assumptions.
The client then uses that text as part of the model input.