This is a Model Context Protocol (MCP) server built with Zuplo that demonstrates how to implement MCP tools and prompts using an OpenAPI-defined API.
About this example
This project implements a Bookmark Manager with MCP integration that showcases:
MCP Tools
list-bookmarks- List saved bookmarks with optional tag filteringsave-bookmark- Save a new bookmark with URL, title, and tagsdelete-bookmark- Delete a bookmark by ID
MCP Prompts
research-roundup- An AI prompt that analyzes recent bookmarks to identify research patterns, themes, and suggest next steps
Key Components
- OpenAPI-based Routes (
config/routes.oas.json) - API routes with MCP annotations - Research Prompt (
modules/research-roundup-prompt.ts) - Custom MCP prompt implementation
Prerequisites
- A Zuplo account. You can sign up for free.
Getting Started
Locally
Working locally is the best way to explore and understand the code for this example. You can get a local version by using the Zuplo CLI:
Once you have the code on your local machine, start the development server:
The API will be available at http://localhost:9000.
You can access:
- API Gateway -
http://localhost:9000 - Local API Route Designer -
http://localhost:9100 - MCP Endpoint -
http://localhost:9000/mcp(POST)
The dev server will automatically reload when you modify:
- Route definitions in
config/routes.oas.json - Handler modules in
modules/
Deploying with Zuplo CLI
Install the Zuplo CLI
Login to Zuplo
Create a New Project
Follow the prompts to create a new project in your Zuplo account.
Deploy to Zuplo
Deploy to a working copy environment:
Using the MCP Server
Once deployed, you can connect to your MCP server using any MCP testing tool:
Model Context Protocol Inspector
MCPJam
You can then test the MCP server locally, or point your inspector at a deployed version.
Local MCP Server
Deployed
The server will expose:
- 3 tools for managing bookmarks
- 1 prompt for analyzing your research patterns