@jaypie/mcp

Prerequisites: npm install @jaypie/mcp

Status: Experimental - APIs may change

Overview

@jaypie/mcp provides utilities for building MCP (Model Context Protocol) servers that expose Jaypie service handlers as tools for AI assistants.

Installation

npm install @jaypie/mcp

Quick Reference

Exports

Export	Purpose
registerMcpTool	Register service handler as MCP tool
createMcpServer	Create MCP server instance

MCP Tool Registration

Basic Registration

import { registerMcpTool } from "@jaypie/mcp";
import { fabricService } from "@jaypie/fabric";

const getUserHandler = fabricService({
  alias: "get_user",
  description: "Get user by ID",
  input: {
    userId: { type: String },
  },
  service: async ({ userId }) => {
    const user = await db.users.findById(userId);
    return user;
  },
});

registerMcpTool(server, getUserHandler);

With Callbacks

registerMcpTool(server, handler, {
  onMessage: (message) => {
    console.log("Progress:", message);
  },
  onComplete: (result) => {
    console.log("Complete:", result);
  },
  onError: (error) => {
    console.error("Error:", error);
  },
});

Service Handler to MCP Tool

Service handlers automatically convert to MCP tools:

Fabric Type	JSON Schema Type
string	string
number	number
boolean	boolean
array	array
object	object
date	string (format: date-time)
uuid	string (format: uuid)

Creating MCP Server

import { createMcpServer, registerMcpTool } from "@jaypie/mcp";
import { handlers } from "./handlers.js";

const server = createMcpServer({
  name: "my-service",
  version: "1.0.0",
});

// Register all handlers
for (const handler of handlers) {
  registerMcpTool(server, handler);
}

// Start server
server.listen();

Response Format

MCP tools return responses as text content:

// Handler returns:
{ name: "Alice", email: "alice@example.com" }

// MCP response:
{
  content: [
    {
      type: "text",
      text: '{"name":"Alice","email":"alice@example.com"}'
    }
  ]
}

Error Handling

Jaypie errors are converted to MCP error responses:

const handler = serviceHandler({
  name: "get_user",
  handler: async ({ userId }) => {
    const user = await db.users.findById(userId);
    if (!user) {
      throw NotFoundError("User not found");
    }
    return user;
  },
});

// MCP receives error with status and message

Input Validation

Input validation from fabric is enforced:

const handler = fabricService({
  input: {
    email: {
      type: String,
      validate: (value) => value.includes("@"),
    },
  },
  service: async ({ email }) => {
    // email is guaranteed valid
  },
});

Prompts and Resources

List Prompts

import { mcp__jaypie__list_prompts } from "@jaypie/mcp";

const prompts = await mcp__jaypie__list_prompts();
// Returns available Jaypie development prompts

Read Prompt

import { mcp__jaypie__read_prompt } from "@jaypie/mcp";

const content = await mcp__jaypie__read_prompt({
  filename: "Development_Process.md",
});

Datadog Integration

Built-in tools for Datadog monitoring:

import {
  mcp__jaypie__datadog_logs,
  mcp__jaypie__datadog_monitors,
  mcp__jaypie__datadog_synthetics,
} from "@jaypie/mcp";

// Query logs
const logs = await mcp__jaypie__datadog_logs({
  query: "status:error",
  from: "now-1h",
});

// Check monitors
const monitors = await mcp__jaypie__datadog_monitors({
  status: ["Alert", "Warn"],
});

LLM Debug Tools

Debug LLM API calls:

import { mcp__jaypie__llm_debug_call } from "@jaypie/mcp";

const response = await mcp__jaypie__llm_debug_call({
  provider: "anthropic",
  message: "Test message",
  model: "claude-sonnet-4",
});

Configuration

Environment variables:

Variable	Purpose
DD_API_KEY	Datadog API key
DD_APP_KEY	Datadog app key
DD_ENV	Default environment filter
DD_SERVICE	Default service filter

Related

• Fabric - Service handler patterns
• LLM Integration - LLM tools
• @jaypie/llm - LLM package