Dynamic Arguments — Prompts

Parametrizing Context & User Interaction Forms

MCP Prompts achieve their full potential when parameterized with strongly typed arguments. Rather than requiring users to manually paste boilerplate parameters or system rules into the chat box, an MCP server exposes argument schemas via prompts/list.

import { z } from "zod";
import { McpServer } from "@modelcontextprotocol/server";

const server = new McpServer({ name: "analytics-prompts", version: "1.0.0" });

server.registerPrompt(
  "generate_quarterly_audit",
  {
    title: "Quarterly Audit Report",
    description: "Generate a quarterly audit template for a department",
    argsSchema: {
      department: z.enum(["Engineering", "Security", "Finance"]).describe("Target business department"),
      fiscalYear: z.string().describe("Four-digit fiscal year, e.g. 2026"),
      includeRemediation: z.enum(["true", "false"]).optional().describe("Include automated remediation playbook")
    }
  },
  async ({ department, fiscalYear, includeRemediation }) => {
    return {
      description: "Executive audit report for " + department + " (FY" + fiscalYear + ")",
      messages: [{
        role: "user",
        content: {
          type: "text",
          text: "Generate a rigorous enterprise audit for " + department + " during FY" + fiscalYear + ". Include remediation plans: " + includeRemediation + "."
        }
      }]
    };
  }
);