MCP — Model Context Protocol — is the open standard that is quickly becoming the infrastructure layer underneath almost every serious AI tool you will encounter in 2026. It's one of the main reasons that AI is shifting from something you consult to something that acts on your behalf. And like most big developments in this space, it has arrived with both significant opportunity and risks.

In this episode, Kyle and Jess do a full deep dive into what MCP is, why the whole industry has moved on it faster than almost any standard in modern tech, and what the upside looks like for a small business that has never had access to serious AI integrations before. We also cover the cons including some new security risks.

This is part one of two. This week we're tackling what it is, the pros, the cons and some quick wins to make sure you understand what you are dealing with before next week's episode gets into the practical setup, the safety framework, and Kyle's actual tech stack. If you are going to connect AI to your business systems — and increasingly, you will be — this is the episode to start with.

What You Will Learn

What MCP is
How MCP differs from APIs, and why that distinction matters
Why OpenAI, Google, and Microsoft all adopted a competitor's open standard within six months
Why agentic AI only delivers on its promise if the AI can move fluidly across multiple systems
The real business advantages: cost efficiency, flexibility, the ecosystem of ready-made connections, and why a cheaper model with good connections beats an expensive one working blind
The risks that matter: over-permissioned access, supply chain vulnerabilities, and a novel attack type called tool poisonin
Some practical rules for staying safe with MCP before next week's full setup guide

Try Granola

If you've ever sat in a meeting, taken what felt like decent notes, and then opened them afterwards and they didn't capture anything, Granola is the tool for you. It runs in the background, captures everything, and turns your notes into something you can actually use. Both Jess and Kyle use it, it plays really nicely with Claude, and it is one of our most highly recommended tools

New users get 100% off their first month using our link: granola.ai?via=early-adoptr

Timestamps:

00:00 What We've Been Up To

06:41 What Is Model Context Protol (MCP) and Why Is Everyone Suddenly Talking About It

17:34 Why MCP Is the Missing Piece for AI That Actually Does Things

22:40 The Real Advantages of MCP for Small Businesses

24:07 The Importance of Your Tool Integrations

27:05 Competitive Advantage through Connected Workflows

29:31 Pros of MCP

31:22 The Downsides to MCP

39:31 Best Practices for Safe MCP Implementation

42:26 AI News: Meta Acquires Molt Book