Last week I broke down what MCP actually is. This week I want to tackle one of t
By Melissa Burdick · May 28, 2026 · Curated by George's Blog
Last week I broke down what MCP actually is. This week I want to tackle one of the most common points of confusion that keeps coming up in conversations about it.
MCP vs AI agents. These are two different things, and knowing the difference will change how you think about both.
An AI agent is where the intelligence lives. It reasons, it plans, it decides, it acts. When you hear about AI doing something autonomously inside a workflow, that's the agent.
MCP is how that agent connects to your world. Your campaign data, your reporting platforms, your workflow tools. Without MCP, an agent is capable but cut off. With MCP, it can reach live data and actually get things done.
John Feng, Pacvue's SVP of Engineering, puts it like this:
"MCP is a resource for agents, not an agent itself. It delivers data or enables action. But without an agent calling on it, it doesn't do anything."
This is why the two get conflated. In practice they work together so seamlessly it feels like one thing. But understanding the difference matters when you are evaluating vendors, building workflows, or asking the right questions of your team.
The question worth asking isn't just "Do we have effective AI?" It's also "Does our AI have what it needs to actually be effective?" That second question is exactly what MCP helps solve for.
More to come next week!