Prompts Overview¶

MCP Prompts are reusable message templates that guide the LLM's behaviour. Unlike tools (which do things), prompts shape how Claude thinks and responds.

How Prompts Differ from Tools¶

	Tools	Prompts
Purpose	Take action, return data	Shape LLM response style
Returns	Any Python type	A string message template
Invoked by	LLM autonomously	User or client explicitly
Side effects	Yes (files, APIs, etc.)	No

Available Prompts¶

Prompt	Parameters	What it does
`analyze_topic`	`topic: str`	Deep structured analysis of any subject
`historical_report`	`topic: str`, `number_of_paragraphs: int`	Structured history report with timeline
`explain_weather_concept`	`concept: str`	Plain-language meteorology explanation

How to Use Prompts in Claude¶

In the MCP Inspector, select the Prompts tab and choose a prompt.
In Claude Desktop, you can invoke them via the slash command or by asking Claude directly:

Use the historical_report prompt for "The Internet" with 4 paragraphs

Apply the analyze_topic prompt to "Large Language Models"

Use explain_weather_concept for "El Niño"

See all built-in prompts