Part of the People Ops AI Brain — ← Back to Hub

<aside> ⚡

TL;DR: Connected context is the single biggest differentiator between AI that feels like a toy and AI that operates as infrastructure. MCP connects Claude to your tools so it can read, write, and search your actual data. Without this, you're copying and pasting forever. With it, you're building real systems.

Read time: 15 min | Delegate to: IT or technical resource | Requires: Some technical comfort

</aside>

<aside> 🔌

This is one of the most strategically important pages in the Brain. Connectors change, tools evolve, new ones emerge constantly. But the principle is permanent: AI that can't touch your actual systems is just a fancy text generator. Connected AI is where the real operating leverage lives. This page helps you think about integration architecture, not just setup steps.

</aside>

What MCP Actually Is

Model Context Protocol (MCP) is a way for AI assistants like Claude to securely connect to external tools and data sources. Instead of copying and pasting information into Claude, MCP lets Claude directly:

💡 The mental model: Think of MCP as giving Claude "hands" to work with your actual tools, rather than just "eyes" to read what you paste in.

Why This Changes Everything for People Teams

Here's a real scenario. A manager Slacks you at 10am: "One of my reports has asked about our shared parental leave policy. Can you send me the details?"

Without MCP

  1. Open the handbook in Notion/Google Docs
  2. Search for parental leave (find 3 different pages, which one's current?)
  3. Copy the relevant section
  4. Paste into Claude: "Can you summarise this for a manager?"
  5. Get the summary, but realise it doesn't mention the recent policy update from last month
  6. Go back, find the update, paste that in too
  7. Get a revised summary
  8. Copy into Slack, reformat
  9. Time: 15-20 minutes

With MCP

  1. Tell Claude: "Find our current shared parental leave policy in Notion and draft a manager-friendly summary"
  2. Claude reads the current policy directly — including last month's update
  3. Claude drafts the summary in the right tone
  4. You review, tweak one line, send to the manager
  5. Time: 2 minutes

The difference isn't convenience; it's capability. Connected AI can do things disconnected AI simply cannot - like cross-referencing multiple documents, checking for consistency, and always working from the latest version. This is the layer that separates "we use AI sometimes" from "AI is part of how we operate."

What Connected AI Actually Unlocks

These are real workflows that MCP makes possible. None of them work with copy-paste.

<aside> 📋

1. Policy consistency checker

You've just updated your parental leave policy. Ask Claude: "Read our updated parental leave policy, then check the employee handbook, offer letter template, and benefits FAQ for anything that now contradicts it."

Claude reads all four documents directly, flags inconsistencies, and drafts fixes. What used to take an afternoon of careful cross-referencing takes 5 minutes.

</aside>