Create xmcp Prompt

Step 1: Gather Information

1. Prompt name (if not provided): What should the prompt be named? (Use kebab-case)
2. Prompt type: Ask which type of prompt they need:
   • Simple - Static prompt text without parameters
   • Parameterized - Prompt with user-provided parameters (most common)
   • Multi-content - Returns multiple content blocks (text, images, etc.)
3. Parameters (if parameterized): What inputs should the prompt accept?
   • Parameter name
   • Type (string, number, enum)
   • Description
   • Whether it's optional
   • Any validation or default values
4. Role: What role should the prompt assume?
   • user - Prompt from user perspective (most common)
   • assistant - Prompt as assistant response template
5. Use case: What is the prompt designed for?
   • Code review
   • Documentation generation
   • Data analysis
   • Content creation
   • Other specific task

Step 2: Generate the Prompt

src/prompts/

File Location

src/prompts/{prompt-name}.ts

Prompt Structure Reference

// 1. Schema (optional) - Define parameters with Zod
export const schema = { /* ... */ };

// 2. Metadata (optional) - Prompt configuration
export const metadata: PromptMetadata = { /* ... */ };

// 3. Handler (required) - Default export function
export default function handler(params?) { /* ... */ }

Quick Reference

Essential Imports

import { type PromptMetadata } from "xmcp";

// For parameterized prompts
import { z } from "zod";
import { type InferSchema, type PromptMetadata } from "xmcp";

Metadata Fields

name

title

description

role

Handler Return Types

string

{ type: "text", text: string }

{ type: "image", data: base64, mimeType: string }

Array<Content>

Use Cases for Server-Exposed Prompts

1. Standardized workflows - Consistent code review, documentation patterns
2. Complex templates - Multi-step analysis prompts with structure
3. Domain expertise - Specialized prompts for specific domains
4. Reusable patterns - Common tasks that benefit from parameterization
5. Client integration - Prompts that MCP clients can discover and use

Detailed Templates

• Simple prompt patterns
• Parameterized prompts with Zod schemas
• Multi-role conversation prompts
• Autocompletion support
• Code review and documentation examples

references/patterns.md

Checklist After Generation

1. File created in

   src/prompts/{prompt-name}.ts

2. If using metadata, ensure it has

   name

   ,

   title

   , and

   description

3. Schema uses

   .describe()

   for all parameters (if parameterized)
4. Handler returns appropriate content format
5. Role is set appropriately for the use case

pnpm build