Build an MCP Product with AI
Emmanuel Paraskakis
Founder, Level 250 | 3x VP, API Products
Your product needs an MCP server for agents to use it. Build it in 3 hours.
Intro price: $249 until April 21st! Use code BMP29.
Every AI agent needs a way to reach your product. MCP is how they do it. Claude, Cursor, ChatGPT, Claude Code, all speak MCP. If your product doesn't have an MCP server, agents can't use it.
But most teams don't know where to start. What tools should you create? How do you name and describe them so agents actually pick the right one? How do you handle auth, errors, and testing? How do you publish MCP so agents can find it?
In this workshop, you'll build two MCP servers from scratch using AI. One from a real API using Zuplo and one from just an idea with Replit. You'll then test them with MCPJam, and see how LLMs and agents use them live.
The Eval loop ties it all together: generate test cases, run them against real models, establish a baseline, fix what fails, and re-test until agents pick the right tool every time.
We also cover authentication, security, and token costs: the real-world challenges you'll hit when your MCP server goes to production.
Walk in curious about MCP. Walk out with two published servers and the skills to build the next one.
Workshop agenda
- 10:00AM EDT
Optional: Introductions
Join early to introduce yourself and meet participants before the workshop starts.
- 10:30AM EDT
MCP in Practice - When to Use and Why
How MCP works, what agents expect, and when to use MCP vs API vs CLI. You'll see a live MCP query in Claude.
- 11:00AM EDT
MCP Tools and Their Design
MCP primitives, the tool lifecycle, and why metadata is agent UX. Write tool names and descriptions that agents pick correctly. Hands-on naming exercise.
- 11:30AM EDT
Build an MCP Server from an API
Take a real API, import its OpenAPI spec into Zuplo, and generate a working MCP server. Connect it to Claude and test it live.
- 12:00PM EDT
Build an MCP Server to Serve Experiences
Start from just an idea and build an MCP server from scratch using Replit. Validate it with MCP Debugger. No coding required - the agent builds it.
- 12:30PM EDT
How to Eval MCP - The Key to Usability
Generate test cases with MCPJam. Run evals against real models. Establish a baseline, identify failures, fix descriptions, and re-test. The eval loop.
- 1:00PM EDT
Challenges: Auth, Security, and Token Costs
The two-hop auth problem, prompt injection risks, and why MCP eats tokens for breakfast. Proposed solutions for each.
- 1:30PM EDT
Optional: Q&A and Next Steps
Extended Q&A and help with your specific product. Path to further learning.
Learn directly from Emmanuel
Emmanuel Paraskakis
Founder, Level 250 | 3x VP, API Products
Who this workshop is for
API and Platform PMs who need to decide when MCP is the right choice and scope their first server for their product
Engineers and Architects who want to rapidly build an MCP server to start testing tool calls
Technical Founders whose product needs to be discoverable by AI agents and must understand the UX and discoverability of MCP
What's included
Live sessions
Learn directly from Emmanuel Paraskakis in a real-time, interactive format.
Lifetime access
Go back to course content and recordings whenever you need to.
Community of peers
Stay accountable and share insights with like-minded professionals.
Certificate of completion
Share your new skills with your employer or on LinkedIn.
Maven Guarantee
Your purchase is backed by the Maven Guarantee.
Frequently asked questions
$499
USD