Claude Comes to You: Extensions & Connectors
So far, every door we've walked through has worked the same way: you go to Claude. You open the app, the desktop window, the terminal — you show up at its place and start talking.
Here's the flip. Claude can also show up at yours.
Claude can come to your apps
Think of a genuinely useful colleague. Most of the time you'd walk over to their desk to ask something. But some days they pull up a chair next to you — at your screen, looking at what you're already working on. That's the idea here. Claude can sit down inside the apps you already have open.
A few places it can do that today:
Your browser. Claude for Chrome is an extension that can actually see the page you're on and act on it — answer questions about it, pull out the bits you need, click through a task for you. It's genuinely handy, and it's also still early. It's a beta feature, and Anthropic is upfront that letting something read and act on whatever's on your screen carries real risk — a sketchy website could try to trick it. So it's worth trying, with a little care about which sites and data you let it touch.
Your team chat. If your workday lives in Slack, Claude can live there too — draft a message, do a quick bit of research, help you prep for a meeting, all without you leaving the conversation.
Your Office apps. Claude can work right inside Word, Excel, PowerPoint, and Outlook — editing the actual file in front of you and carrying what it knows from one app into the next, so you're not re-explaining yourself when you jump from a spreadsheet to an email.
Notice what didn't change. It's still one Claude, the same intelligence we've been meeting all along — just wearing different work clothes, showing up in a new room.
And you can give it more to work with
The second half of this lesson is about making Claude better at your specific stuff. Two ideas do most of the work, and they're easy to keep straight.
Skills are like handing Claude a recipe. Say there's a task you do the same way every time — formatting a weekly report, filling in a certain kind of form. Instead of re-explaining your way each time, you give Claude a Skill: a saved set of instructions for how you do that job. Then it follows the recipe and gives you the same solid result every time, no reminding needed.
Connectors are the plugs. They let Claude reach into the other apps you use — your calendar, your documents, your tickets — so it can answer using your actual stuff instead of guessing. Back in the Cowork lesson, we said an agent is just an LLM with hands. Connectors are some of those hands, reaching into your tools.
There's a shared standard underneath that makes connectors possible, called MCP — think of it as a universal plug shape, so all sorts of apps can connect the same way. You don't need to learn anything about it; just know the word exists, because you'll see it around.
(You may also bump into Plugins — bundles that package up capabilities to share, mostly over in Claude Code. File that one under "good to know it exists" for now.)
One clean way to hold all of this
It can start to feel like a pile of features, so here's the sorting hat. Some of these change where Claude shows up. Others change what it can reach and how reliably it does your specific jobs.
- Surfaces — browser, Slack, Office — are about where you meet Claude.
- Skills are about teaching Claude to do a task your way.
- Connectors are about plugging Claude into your other tools.
We've now toured the doors and the ways Claude reaches beyond them. But every one of these — the app, the browser extension, the Skill quietly following your recipe — runs on the same thing underneath: the models, the actual engine doing the thinking. That's where we head next, and it's the piece that finally explains why there's more than one "Claude" to choose from.