---
title: "Connectors: One Click to Your Data"
description: "Drive, Gmail, Calendar — live-data answers with no terminal."
order: 19
duration: "10 min"
chapter: "05-trusting-your-own-data"
type: lesson
---

## MCP without a terminal

Here is the part most people miss: you are probably already one click away from your first MCP connection, because on claude.ai and in the desktop app, MCP servers wear a friendlier name — **Connectors**.

Open Settings → Connectors and you will find a directory of ready-made connections: Google Drive, Gmail, Google Calendar, Slack, Notion, Linear, and a growing catalog of others. Each one is an MCP server that a provider runs for you. Connecting is an authorization flow — click, sign in, grant access — not a configuration file. Once connected, every conversation (and every Project) can use it, on every device signed into your account.

Sarah connects two:

- **Google Drive** — where the scanned invoice PDFs and supplier contracts live
- **Gmail** — where every vendor conversation, dispute, and delivery notice ends up

Then she asks, in a plain chat: *"Find the emails from Pacific Foods in the last two months about substitutions, and check the Drive folder for the invoices from those weeks. Did we get billed for the original items or the substitutes?"*

Claude searches Gmail, finds three substitution notices, pulls the matching invoice PDFs from Drive, and lines them up: two invoices billed the substitute items correctly; one billed the original (more expensive) item — an $84 discrepancy nobody had caught. This is the trust inversion again, but now across *live systems* instead of an exported CSV. Nothing was exported. Nothing was attached. The data stayed where it lives, and Claude went to it.

Two things to internalize from this exercise:

**Authorization is the control point.** A connector can only see what you granted it — and you can disconnect it in the same settings panel. When your team asks "is this safe?", the honest answer is: it has exactly the access you gave it, it acts as you, and every action is visible in the conversation.

**Connectors are MCP.** The Google Drive connector and the QuickBooks server you will wire up in Chapter 07 are the same species. One has a polished authorization flow; the other you register yourself. Understanding this now means Chapter 07 will feel like more of the same, not a new subject.

<div class="exercise">
  <div class="callout-label">Try This</div>
  <p>Open Settings → Connectors (claude.ai or the desktop app). Connect ONE system where your real work lives — Drive, Gmail, Calendar, Notion, Slack. Then ask a question that spans it: "Summarize my meetings this week and pull up the documents related to each." Or the Cornwall pattern: cross-reference two sources and look for a discrepancy.</p>
  <p>Then review what just happened with a critical eye: What did Claude actually access? Could you trace each claim back to a source message or file? That audit habit is what makes connected AI deployable in a real organization.</p>
</div>

## Check your understanding

Answer in your own words — write it down before opening the key. Your tutor grades against the criteria and generates fresh variants on retries.

**Q1.** "Should we build a Claude integration for Google Drive?" a colleague asks. What is the actual relationship between Connectors and MCP, and what does that mean for their question?

<details>
<summary>Answer key — attempt every question first</summary>

## Answer key

### Q1

**Model answer:** Connectors ARE MCP — hosted MCP servers with a polished authorization flow. So the integration already exists: connect Google Drive in Settings → Connectors with one authorization, and it works across every surface and device. Build a custom MCP server only for systems nobody hosts yet.

**Pass criteria:** identifies connectors as hosted MCP, not a different technology; conclusion: authorize, don't build, for common services

</details>


**Next:** [What MCP Is](/mastering-claude/05-trusting-your-own-data/20-what-mcp-is/)