Claude Code Explained: MCP, Skills, Plugins and Routines for Beginners

But, if you only use Claude Code to answer questions and reply to emails, you're using about 10% of it. The other 90% is hiding behind words like MCP, plugins, and slash command. So, in this video, I want to show you how to get the best value out of your current Claude subscription. And we'll go over all of these terms, what they mean, how to implement them, and why they're useful. Let's get started.

And the first thing that I want to set up with you are connectors, because they're one of the easiest things to set up, and I feel like they make one of the biggest differences straight away. So, think of connectors like giving your new employee access to the tools that you use. Basically, you're giving Claude permission to go into your Gmail, calendar, Canva, ClickUp, and that means when you get a customer email asking what time can they meet in Chicago time, and you're based in London, it can actually answer that question and draft an email that's right inside your Gmail account without having to copy it from Claude and pasting it into Gmail. To access connectors, you just need to go to settings and connectors. And then, you'll see the ones that you have already set up, if any, and there's a discover page which has a lot of the other ones.

And all you need to do is just click connect and log into your account. There are a few ones that I would set up in the first place, and that's Google Calendar, Google Drive, and Gmail account, because that gives Claude a lot of information about you, what you're currently working on, and what are your priorities. And then, I would go into tools like Canva, Adobe, or DaVinci Resolve. Whatever it is that you use to edit images or make content. I'll be honest though, I started with the Canva connector, but now I actually mostly let Claude do the image and video editing by itself.

And either use just Claude Code or I use Claude Code connected to Hixfeld. Next up, I would add a tool like Granola or Fathom. So, whenever you have a sales call, the transcript from that meeting is recorded and the follow-up is done immediately. And what is great about those tools as well is that you can actually get coaching from them. Because if Claude has information about the last 10 or 15 calls that you have made, it's able to show you the patterns and it's able to let you know what are some of the things that you could have done better during those calls, which is super useful.

When you are connecting those apps, I would remember about two things. The first one is read the permission screen. So, because when you click connect, the app will tell you what Claude will be able to do. Like read, write, or delete files. So, have a look at it say yes.

And two, I would ask for a draft first. Like saying draft a reply instead of reply or draft a post instead of posting it. So, that you can check it before it goes out. And once you trust it, you can change those permissions and loosen it up. Next up, we've got MCP.

And it stands for Model Context Protocol. And you can forget what it means immediately. Think of it like universal plug socket. Before it existed, every app needed its own special adapter in order to talk to AI. MCP is like one standard plug that everyone can use.

And here's the thing, a connector is just an MCP server that someone has made really easy to switch on. So, you have already been using MCP without knowing it if you have ever connected any of those apps. But knowing what MCP is and how you can set it up is actually really useful, because there are a lot of tools that are not on that connector list. And companies are now creating those MCP server links on their website that you can just hand that link in to Claude, and Claude is able to set it up for you. But, some tools don't have a connector or an MCP server.

And that's where API keys come in. And API keys, think of API keys as a password that lets one piece of software use another. Lots of tools like 11 Labs or Open AI have a page where you can click create an API key. And you get a long string of these letters and numbers. But, there is a bit that you really need to understand.

An API key is not just a password to your account. It's a password with your credit card attached to it. Whoever has that API key can use your account as if they were you and bill you. And unlike when you use a normal login process, where there is a two-factor verification, so you get, for example, an email and you have to paste that code into the website, if someone has an API key, the door is open. So, you need to treat it exactly like you would treat your bank password.

And I've heard so many stories of API keys being leaked by accident, where someone created a screenshot or a screen recording or a project that they shared online. And the thing is, there are bots whose whole job is to scan the internet, especially GitHub, looking for these keys and exploiting them. So, if you post one publicly, it can be found and used within minutes. So, before you share anything, be extra careful. And if a key ever does slip out, don't panic, just go to that website, delete that key, and make a new one.

The old one stops working straight away. The other thing you have to remember is you should never paste your key into the chat with Claude. And unfortunately, Claude often says, "I need your API key." And it feels like the obvious thing to do. Because the thing is that Claude doesn't actually need to see your key before using it.

It just needs to know where it's kept, which brings me to how to do it properly. Claude can create something called a dot ENV file, which is basically a little drawer for your keys inside your project. You open it yourself, paste the key in, and save it. And now Claude can use that key without the key actually ending up in the chat. It kind of works like Claude can use a key when it runs something, but it can't open that drawer and take a look.

An even safer way to use API keys is using GitHub, which has a special lock box called secrets. And if you've got running on GitHub, like an automation, you put the key in there instead of in your files. And once you save it, even you can't see it again. GitHub just hands it to your automation whenever it runs. Next, let's talk skills.

And this is where Claude Code starts to feel like it actually knows how you work. Because think of a skill like a recipe card. It's a little file that says, "When Wiktoria asks for an Instagram thumbnail, here's exactly how she likes it done. These are the fonts that she likes using. This is the hand light style.

This is the tool that she uses. Save it in whatever folder." And Claude can read that recipe whenever it needs it. So, you can have loads of them, and they don't get in the way. Easiest way to make one is do a task once, get it to how you like it, and then say, "Turn what we just did into a skill, so you can do it in the same way next time.

Now, let's talk about plugins. Think of plugins like a gift box. Inside, there might be a few skills, a few slash commands, and some MCP connections. And they're all bundled together by someone who has figured out how to do a certain task. For example, a marketing plugin, a sales plugin, a design plugin, etc.

And if you're just starting with plugins and skills, here are the ones that I found to be most useful as business owner. The first one is marketing one, where it can plan a whole campaign with a content calendar and draft posts and emails for each platform. The next one being sales. And this one is useful if you sell anything using calls, proposals, follow-ups, etc. But where can you actually find plugins?

So, there are lots of official plugins that are already provided by Claude. And in order to access them, you just click settings, and then there is a tab with the plugins. And you can browse the ones that are there. The ones that I found to be really useful are the marketing sales ones and a PDF ones, which allows Claude to quickly scan a PDF and extract information from it. I would be cautious about installing plugins from the internet unless they have very high rating on GitHub.

Because a plugin is like someone else's recipe that's running on your computer, sometimes with their own connections. So, stick to the ones from the sources that you trust. And a great thing about plugins and skills is the fact that you don't actually need to mention a specific name of that plugin. If, let's say you say, "Oh, plan my content for January," it will know that you have a marketing plug and Claude will use that plugin or that skill in order to do the task. And this brings us to slash commands.

Think of slash commands as shortcuts. They are little buttons that you press instead of typing a whole sentence. You just need to type a forward slash in the chat box and a sort of menu pops up. And there are loads of them. So, let me show you the ones that I actually use.

The first one is slash init, which means that Claude looks around your folder and writes itself a Claude.md file, which is a briefing note that it reads at the start of every chat. Slash clear is one that wipes the chat and starts fresh, which is handy when you switch to a new task. Slash compact squashes a long chat down into a summary, so you can keep going with the task. Slash context show you how full Claude's memory is right now.

Slash rewind takes the conversation and Claude changes back to an earlier point. It's like an undo button. And every skill that you make actually becomes your own slash commands, too. So, my thumbnail skill is just slash YouTube thumbnails. Now, let's talk about Claude Code.

So, in Claude Code, a project is a folder. And everything that I've mentioned so far can live inside it. So, its own Claude.md briefing, its own .env file with its own keys, its own MCP servers, its own settings.

So, when I open my podcast folder, Claude knows all about my podcast. When I open my proposals folder, it knows about the proposals. And this is actually how my AI team works. I have Alina, who's my chief of staff, and she has her own folder. Iris, my social media editor, has their own.

And just so you're not confused, there's something called projects in the normal Claude chat app, but that's just for chat. In Claude Code, a project is a folder. Now, let's talk about artifacts. And here's the easiest way to think about them. Imagine that you ask your assistant for this month's sales numbers, and they could read them out to you, or they could hand you a nice one-page report with a chart on it, and all of the figures drawn in a visual way.

Normally, Claude does the first one, where you get a wall of text in the chat, and an artifact is the second one. With artifacts, Claude allows you to make a little web page instead of giving you a wall of text. And you can click that link, and it opens like any website, but it's done locally, and you can scroll it, interact with it, click buttons, and tick things off. And to create one, all you have to do is add make an artifact to the end of any request. So, what do you actually use it for?

Well, anything that you would rather look at than read. So, for example, a dashboard of your Instagram numbers, or a checklist for onboarding a new client, or a one-page summary to send a client after a call. The one I created myself is review desk, which is when Claude edits a reel for me or a video. Instead of uh opening that video and saying, "Oh, at about 20 seconds in, uh delete that frame." I can just Claude just gives me the page, and I watch the video right there, click at the exact second that I feel like something is off, I type my note in, and Claude can read my notes and fix them.

And they're all private. Only you can see an artifact until you decide to share the link with someone. And the last bit is routines. And I feel like this is where it starts to feel a bit like magic. A routine is a job that Claude can run on a schedule in the cloud.

So, it happens even if your laptop is closed. And this is where everything from this video really comes together, because a routine can use your connectors, your skills, and your plugins. So, it can read your inbox, check your calendar, look at your meeting notes, and use your recipe, all on its own. So, and here are some of the most useful routines I have set up as business owner. The first one is Monday morning briefing.

Where every Monday at 7:00, I have it look at my calendar and inbox for the week and tell me three priorities, which meetings I need to prep for, and anything urgent that I've missed from the previous week. So, I start the week knowing exactly where I stand. The second one is meeting follow-up, which means that at 6:00 p.m., Claude looks at my recordings with each client and drafts a follow-up email with what we agreed on and all of the next steps.

And the third one is a weekly content report, where every Tuesday Claude goes through my YouTube analytics, Instagram analytics, and it looks at that data and tells me what are some of the videos I've done really well, and what are some of the ones that not, and it suggests ideas for improvement. I hope you give at least a few of these a go, and if you want a cheat sheet with every prompt from this video, it's linked in the description. And if you can subscribe to this channel, it would mean the world to me. I looked at the analytics and only 96% of people viewing these videos are subscribed to the channel, so it would be incredible if you are one of those people and you subscribe to the channel.