What is MCP? A non-technical guide for HR leaders

Model Context Protocol. 

Sounds like something only a senior engineer could understand. But, in practice, it’s one of the most approachable uses of AI. Without writing a single line of code, you can securely link your AI application to employee data. 

And using your version of Claude or ChatGPT with live data opens up time-saving opportunities. You can use AI to calculate comp ratio, check vacation balances, analyze payroll runs, build PTO dashboards, draft employment verification letters, and more.

Let’s take a look at MCP from a non-technical lens. 

What is MCP?

Definition: Model Context Protocol (MCP) is a set of rules for how an AI application asks a piece of software for information, and how the system responds.

Imagine you have a storage room full of paper employee files. Manila envelopes and bankers’ boxes. To safeguard the files, your company establishes a process where every time you want to access info, you have to fill out a form and hand it over to a file clerk. The clerk reviews your form and, if accurate, brings you the files you requested. 

MCP works the same way. Only digital and instant. 

And if we break down the name Model Context Protocol, we’ll see every part at work: 

• Model. The AI app you already use, like Claude or ChatGPT. It knows a lot, but can’t access your HR data without going through a predefined process.

  
  

• Context. Your data, like headcount, leave balances, compensation, payroll history. Without it, even the most robust AI model will give you generic results.
  
  

• Protocol. The agreed process for asking and answering. Using it, any compatible and approved AI app can connect to data.

Without an MCP, or another protocol for accessing data like an open API or CLI, your data just gathers dust in the digital equivalent of a storage room. It’s there, ready to be used, but your AI tools can’t reach it. 

And before we go further, there’s an important caveat. Your HR platform has to proactively run an MCP server and open it to you. Many don’t, keeping your data locked away and unavailable to AI applications. 

In practice

Meet Hannah, an HR leader at a 100-person company. She’s AI-savvy. She uses Claude every day to draft policy, create offer letters, and clean up messy spreadsheets. 

But Hannah has a problem. Claude can’t see her HR data. So she has to copy rows from a spreadsheet, paste them into the chat, and ask her question. Or she’s stuck using hypotheticals. "Assume a comp band midpoint of $95,000."

So she connects Claude to Plane through Plane’s MCP server. To do this, she simply adds Plane as a connector in Claude’s settings and signs in. Setup takes a few minutes, once, and doesn’t require an engineer. 

Now, instead of wrestling with static data or assumptions, Hannah just asks Claude questions directly. "What is Priya's comp ratio if we move her to $102,000?" Claude pulls the live record and answers.

Hannah’s AI app finally has access to live data, making it smarter and more useful. 

What changes for HR

Why should an HR leader like Hannah bother using an MCP? In short, when Hannah syncs her AI apps and HR data, new doors open. 

• Data is real-time. Every AI response pulls from the live record. You don’t have to export or copy/paste. You don’t have to worry about whether the data is up to date. 

• Questions get bigger. Before, cross-referencing six payroll runs or breaking down leave by department was a half-day of pivot tables. After MCP is set up, it’s as easy as a one-sentence prompt. 

• Busywork drops off. Verification letters, dashboards, comp calculations. The work that was always urgent and never strategic gets drafted while you do something else because AI can fill in the blanks with your data. 

What stays the same

Does MCP mean giving something up? The short answer: no. The rules that governed your HR data yesterday govern it the same way today.

• Data only goes where you send it. MCP is a secure connection between your HR data and AI application. Note: What your AI application does with the information once it arrives is governed by your agreement with that provider. 

• Permissions still apply. An AI application connected through MCP sees what your access level allows and nothing more. 

• You still review the work. AI applications make mistakes, and questions get misread. You see the output before anything happens with it.