I’m back with Den Delimarsky to talk about the Model Context Protocol (MCP) — the open standard that’s quietly transforming how AI assistants connect to data, tools, and services.

We unpack what MCP is, why it matters, and how it’s becoming the foundation for agentic workflows, developer tools, and authentic AI experiences. Watch the full conversation below:

Besides Youtube, you can listen/watch on Spotify and Apple Podcasts as well if that is your jam!

The TL;DR

The Model Context Protocol is to AI what HTTP was to the web — a universal connector that lets AI assistants securely access real-world data and tools.

It’s open, extensible, and vendor-neutral, designed to help AI work with humans instead of replacing them. MCP provides the “plumbing” for the magic behind next-gen assistants and agentic systems. We dive into this and more, and the following section headings and timestamp can help you jump to any topic that catches your eye.

0:00 – Intro: Setting the Stage
Kirupa introduces Den Delimarsky and outlines the episode’s focus: the Model Context Protocol (MCP), an open standard enabling AI assistants to access real-world data and services.

1:00 – Origins of MCP
Den explains how he became a core maintainer of MCP after contributing improvements to the authorization spec.

2:20 – What Is MCP?
An overview of MCP as the “USB-C for LLMs” — a universal bridge that connects large language models to real-world data, tools, and APIs.

4:30 – Why Not Just Use REST APIs?
Discussion of why existing systems aren’t suited for LLMs: authentication, streaming, and standardization challenges. MCP provides a unifying layer.

7:00 – Local vs Hosted MCP Servers
The two types of MCP servers — local (stdio) and hosted (HTTP) — and how they integrate with tools like Figma.

8:00 – Developer Adoption & Tooling
Kirupa connects MCP to his own work on AI-powered developer tools and the role of the MCP.json configuration.

9:00 – Discoverability & The MCP Registry
Den introduces the MCP Registry, a public index of MCP servers (registry.modelcontextprotocol.io), and how it improves discovery and management.

11:30 – Security & Trust Challenges
They discuss early-stage security issues, dependency trust, and parallels to NPM vulnerabilities.

16:00 – Competing MCP Implementations
On balancing an open standard with competing versions (e.g., Stripe vs PayPal analogies) and preventing fragmentation.

19:30 – Why Neutrality Matters
MCP’s vendor-agnostic design ensures it works across providers (Anthropic, Google, OpenAI, etc.).

21:00 – The Road Ahead for MCP
Den highlights upcoming work on profiles and extensions to make MCP more customizable and enterprise-ready.

25:00 – Avoiding Fragmentation & The Power of Community
How MCP’s open governance prevents protocol “balkanization” and encourages convergence across competing systems.

31:00 – Lessons from History: Standards Win
They compare the rise of MCP to historical format wars (VHS vs Betamax, JSON vs XML) — standards survive when communities unite.

33:00 – Agents and the Coming Paradigm Shift
The conversation expands: MCP as the plumbing for agentic workflows where users issue natural language commands instead of using GUIs.

39:00 – AI Ethics & Authenticity
Den voices concern about corporate bias and authenticity — ensuring agents serve users, not advertisers.

44:00 – Democratization vs AI Slop
They reflect on the flood of low-effort AI content, “AI slop,” and the tension between creativity and convenience.

52:00 – Empowerment Through AI
Den emphasizes using AI to empower humans, not replace them — drawing parallels between art, code, and human creativity.

59:00 – Taste and the Future of Work
Both discuss “taste” as the differentiator in an AI world where anyone can generate apps or art — design sense and judgment will matter most.

1:06:00 – AGI, Creativity, and Human Uniqueness
Philosophical exploration of whether AI can ever replicate human taste or creativity — skepticism prevails.

1:15:00 – SpecKit, Guardrails, and the Future of Coding
Kirupa mentions SpecKit and the need for determinism and guardrails in AI-assisted development.

1:18:00 – Expertise Still Matters
Even in an AI-augmented world, secure and maintainable software will still depend on experienced engineers.

1:21:00 – Agentic Workflows and the SR Agent Future
They discuss automated debugging, observability, and self-healing systems — the rise of AI “SREs.”

1:24:00 – Cutting Through the Hype
Final reflections on distinguishing signal from noise in the AI space, and how MCP fits into the long-term infrastructure of agentic systems.

1:25:30 – Conclusion: Plumbing for the Magic
They wrap up by tying it back to the beginning: MCP is the invisible plumbing that will make AI interactions feel like magic.

Conclusion

MCP shows that progress in AI isn’t just about smarter models — it’s about better connections. By setting a common standard, we’re paving the way for tools that truly work together. Thanks for joining Den and me on this deep dive, and see you in the next episode!

Just a final word before we wrap up. What you've seen here is freshly baked content without added preservatives, artificial intelligence, ads, and algorithm-driven doodads. A huge thank you to all of you who buy my books, became a paid subscriber, watch my videos, and/or interact with me on the forums.

Your support keeps this site going! 😇