MCPPrompt Class API Reference

---

title: "MCPPrompt Class API Reference" description: "Complete API reference for the MCPPrompt base class in mcp-framework — create reusable, parameterized prompt templates for AI models." order: 4 keywords: ["MCPPrompt", "MCP prompt", "mcp-framework prompt", "prompt template", "AI prompts", "reusable prompts"] date: "2026-04-01"

Overview

Prompts are the third MCP primitive alongside tools and resources. They help standardize how AI models approach specific tasks by providing pre-built prompt structures with fillable parameters.

Class Signature

import { MCPPrompt } from "mcp-framework";

class CodeReviewPrompt extends MCPPrompt {
  name = "code_review";
  description = "Generate a thorough code review for a given code snippet";
  arguments = [
    {
      name: "code",
      description: "The code to review",
      required: true,
    },
    {
      name: "language",
      description: "Programming language of the code",
      required: false,
    },
  ];

  async getMessages(args: Record<string, string>) {
    return [
      {
        role: "user" as const,
        content: {
          type: "text" as const,
          text: `Please review the following ${args.language || ""} code:\n\n${args.code}\n\nProvide feedback on: code quality, potential bugs, performance, and best practices.`,
        },
      },
    ];
  }
}

export default CodeReviewPrompt;

Properties

PromptArgument

interface PromptArgument {
  name: string;
  description: string;
  required?: boolean;
}

Methods

getMessages(args)

Returns an array of messages that form the prompt.

async getMessages(args: Record<string, string>): Promise<PromptMessage[]>

Parameters:

• args — Key-value map of argument values provided by the caller

Returns: Array of PromptMessage objects with role and content

PromptMessage Structure

interface PromptMessage {
  role: "user" | "assistant";
  content: {
    type: "text";
    text: string;
  };
}

Complete Examples

Analysis Prompt

import { MCPPrompt } from "mcp-framework";

class AnalyzeDataPrompt extends MCPPrompt {
  name = "analyze_data";
  description = "Analyze a dataset and provide statistical insights, trends, and recommendations";
  arguments = [
    {
      name: "data",
      description: "The dataset to analyze (CSV, JSON, or plain text)",
      required: true,
    },
    {
      name: "focus",
      description: "Specific aspect to focus on (trends, outliers, correlations)",
      required: false,
    },
  ];

  async getMessages(args: Record<string, string>) {
    const focusInstruction = args.focus
      ? `Pay special attention to: ${args.focus}.`
      : "Provide a comprehensive analysis.";

    return [
      {
        role: "user" as const,
        content: {
          type: "text" as const,
          text: `Analyze the following dataset:\n\n${args.data}\n\n${focusInstruction}\n\nPlease provide:\n1. Summary statistics\n2. Key trends\n3. Notable outliers\n4. Actionable recommendations`,
        },
      },
    ];
  }
}

export default AnalyzeDataPrompt;

Multi-Turn Prompt

import { MCPPrompt } from "mcp-framework";

class DebugPrompt extends MCPPrompt {
  name = "debug_error";
  description = "Systematically debug an error with context about the codebase";
  arguments = [
    {
      name: "error",
      description: "The error message or stack trace",
      required: true,
    },
    {
      name: "context",
      description: "Relevant code context or file contents",
      required: true,
    },
  ];

  async getMessages(args: Record<string, string>) {
    return [
      {
        role: "user" as const,
        content: {
          type: "text" as const,
          text: `I'm encountering the following error:\n\n\`\`\`\n${args.error}\n\`\`\`\n\nHere's the relevant code context:\n\n\`\`\`\n${args.context}\n\`\`\``,
        },
      },
      {
        role: "assistant" as const,
        content: {
          type: "text" as const,
          text: "I'll analyze this error systematically. Let me break down the issue:",
        },
      },
      {
        role: "user" as const,
        content: {
          type: "text" as const,
          text: "Please identify the root cause, explain why it happens, and provide a fix with code.",
        },
      },
    ];
  }
}

export default DebugPrompt;

File Placement

Prompts are auto-discovered from the src/prompts/ directory:

CLI Scaffolding

Generate a new prompt:

npx mcp-framework add prompt MyPrompt

This creates src/prompts/MyPrompt.ts with the correct boilerplate. See the CLI Reference for details.