In this one-off interactive, gamified workshop, we’ll simulate real-world work scenarios at your organisation via a board game, helping you identify and eliminate bottlenecks, inefficient processes, and unhelpful feedback loops.
Workshop Details
For the first few months I used Claude seriously, it was brilliant but isolated. It could write, analyze, reason, build — but it couldn't see anything. Not my files, not my calendar, not my email, not my design tool. Every conversation required me to copy-paste context into the chat window like I was feeding a printer one page at a time.
Then I connected Google Drive. Then Gmail. Then Notion, Google Calendar, and Figma.
The change wasn't incremental. It was a phase transition. Claude went from being a smart text box to being something that operates inside my actual work environment. It finds files I forgot I had. It cross-references meeting notes with email threads. It pulls design assets and drops them into presentations. It doesn't just know what I tell it — it knows what my tools know.
The technology that makes this possible is called MCP — the Model Context Protocol. And while the name sounds like something from an enterprise architecture meeting, the idea behind it is surprisingly simple. Most of what I'm covering here comes from the MCP documentation and Anthropic's learning platform, translated into what it actually means for daily work.
The USB-C Analogy (Which Is Actually Perfect)
Anthropic describes MCP as "a USB-C port for AI applications." I normally roll my eyes at tech analogies, but this one is spot on.
Before USB-C, every device had its own proprietary connector. Your phone charger didn't work with your laptop. Your camera cable didn't work with your tablet. Every new device meant a new cable in the drawer.
Before MCP, every AI integration was custom-built. If you wanted Claude to talk to Google Drive, someone had to build that specific connection. Want it to talk to Notion too? That's a separate integration. Slack? Another one. Each tool, each AI application, each connection — built from scratch.
MCP is the universal standard. Build a connector once, and it works with any AI application that supports the protocol. Claude, ChatGPT, VS Code Copilot, Cursor — they all speak MCP. And every tool that publishes an MCP server becomes instantly accessible to all of them.
That's why the ecosystem is growing fast. It's not Anthropic building every integration. It's the tool makers themselves — Notion, Figma, Sentry, Slack — publishing their own MCP servers because the standard makes it worth their while.
What MCP Actually Lets Claude Do
From a user perspective, MCP connectors give Claude three capabilities that change the game:
| Capability | What It Means | Example |
|---|---|---|
| Tools | Claude can perform actions — search, create, update, delete — inside your connected apps. | Search your Drive for a file. Create a Notion page. Draft a calendar event. Send a Slack message. |
| Resources | Claude can read contextual data from your apps — file contents, database records, design specs. | Read your Figma components. Pull a spreadsheet from Drive. Access a Notion database. |
| Prompts | Connectors can provide pre-built interaction templates that structure how Claude engages with specific tools. | A project management connector that includes a template for sprint planning queries. |
The important part: these aren't one-directional. Claude doesn't just read your tools — it acts inside them. When I ask Claude to find a client proposal and update the timeline, it searches Drive, opens the document, and makes the change. When I ask it to check my calendar and draft a scheduling email, it reads the calendar, identifies the free slots, and composes the email with the right context.
That's the difference between an AI that can talk about your work and an AI that can do your work.
My Connected Setup
Here's what I have connected and how I actually use each one:
| Connector | What I Use It For | How Often |
|---|---|---|
| Google Drive | Finding client documents, proposals, past deliverables. Claude searches by content, not just filename — which is more than I can say for Google's own search half the time. | Daily |
| Gmail | Pulling context from email threads for status updates, drafting replies with full thread context, finding specific conversations. | Daily |
| Google Calendar | Checking availability, drafting meeting invites, understanding my schedule context when planning deliverables. | Several times a week |
| Notion | Accessing project databases, checking task status, creating and updating pages for client work and content planning. | Daily |
| Figma | Pulling design references, checking brand assets, reviewing component libraries when building presentations or artifacts. | Weekly |
The compound effect matters more than any single connector. When Claude can search my Drive and read my email and check my Notion databases in the same conversation, it can do things that no individual tool could do alone. "Find the latest version of the client proposal, check if there's any email feedback I haven't addressed, and update the Notion project tracker" — that's a task that spans three tools and would take me fifteen minutes of tab-switching. Claude does it in one prompt.
How to Set It Up (It's Simpler Than You Think)
Connecting tools in Claude is not an engineering project. It's a settings toggle.
Go to Settings in Claude. Look for Connectors (or MCP). You'll see a list of available integrations. Click one, authenticate with your account (standard OAuth — the same "Sign in with Google" flow you've done a thousand times), and it's connected. No API keys, no server configuration, no terminal commands.
For the built-in connectors — Google Drive, Gmail, Calendar, Notion, Figma, Slack — this takes about thirty seconds per tool. You authenticate, grant permissions, and Claude immediately has access.
If you're more technical, MCP also supports custom servers. Claude Code and Claude Desktop can connect to local MCP servers that expose your own databases, file systems, or internal tools. But for most people reading this, the built-in connectors cover the core workflow.
Start with Google Drive and Gmail. Those two alone cover 80% of the value. Add Notion if you use it for project management. Add Calendar if you want scheduling context. Add Figma if you work with design. You can always connect more later — the point is to start.
Why This Matters More Than You Think
Here's the thing that took me a while to internalize: every other feature I've covered in this series gets better when Claude is connected to your tools.
Memory tells Claude who you are. Connectors let it see what you're working on — right now, in real time, across your actual tools.
Projects give Claude context boundaries. Connectors let it pull live data from inside those boundaries instead of relying on static uploaded documents.
Extended Thinking gives Claude time to reason. Connectors give it something real to reason about — your data, your documents, your schedule, your communication history.
Agents give Claude the ability to string together multi-step tasks. Connectors are the hands and eyes that make those steps possible. An agent without connected tools is just a chatbot with ambition.
MCP is the infrastructure layer underneath everything else. It's not the feature you show off in demos. It's the feature that makes all the other features actually work in real life.
The Ecosystem Is Moving Fast
MCP is an open protocol, not an Anthropic-only product. That's a deliberate choice and it's paying off. The list of tools with MCP support grows weekly. Beyond the ones I use, there are connectors for Slack, GitHub, Salesforce, databases, local file systems, and dozens of others — with more appearing constantly.
The architecture is also smarter than it looks from the outside. MCP servers can be remote (hosted by the tool provider, like Notion's official server) or local (running on your machine, accessing local files or databases). They support real-time updates — when something changes in your connected tool, the MCP server can notify Claude without you having to ask. And the protocol handles security properly, using standard OAuth flows rather than asking you to paste API keys into text fields.
For the enterprise and developer crowd: MCP is supported across Claude, ChatGPT, VS Code, Cursor, and a growing list of AI development environments. Build an MCP server for your internal tools once, and every AI application your team uses can connect to it. That's a fundamentally different proposition than building custom integrations for each AI vendor.
The Phase Transition
I keep coming back to that word: phase transition. Not a gradual improvement. A fundamental change in what the tool is.
Claude without connectors is an incredibly smart assistant who works in a sealed room. You have to carry everything to it — every document, every piece of context, every bit of information. It can only work with what you bring.
Claude with connectors is that same assistant, but the walls are gone. It's sitting in your office, at your desk, with access to your filing cabinet, your inbox, your calendar, and your design tools. You don't carry context to it. It goes and gets what it needs.
That's the difference. And once you experience it, going back feels like working with one hand tied behind your back.
MCP is the least glamorous feature in this series and probably the most important one. It's the reason everything else works in practice, not just in demos. Connect your tools. Start with Drive and Gmail. Give Claude access to where your work actually lives. That's when it stops being a chatbot and starts being infrastructure.
