What are APIs and MCP?

APIs let one piece of software ask another to do something, and MCP is an agreed way for AI to use those same connections. Together they mean the programs you already pay for can work as one system instead of several separate ones, with a person moving information between them by hand.

That can sound abstract, so we'll use a real example: what happens in our own business when somebody books a call with us.

What happens when someone books a call with us

When someone books a free call through our booking system, two things happen on their own. We get an email telling us who has booked and when. And the call goes straight into our calendars, so it shows up on our phones and our computers without anyone typing it in.

Nobody copies a name from one screen to another. Nobody forgets to add it to the diary. It's a small thing, but it's exactly the kind of small thing that goes wrong on a busy week, and it runs the same way every time.

Nothing in that setup is AI. It's automation, and it works because of APIs.

What is an API?

API stands for Application Programming Interface. Think of it as a service counter that a piece of software keeps open for other programs. Another program walks up, asks in the agreed format ("add this appointment", "send this email", "give me today's bookings"), and the software either does it or hands back the answer.

Most of the software a small business uses has one: accounts packages, email, calendars, booking systems, online shops, file storage. The counter is how they talk to each other without a person in the middle.

What that means for you is simple. If two programs you use both have an API, they can almost always be connected. Our booking system tells the calendar and the mailbox what's happened through their counters, and they do the rest.

If you're copying the same details between two programs every week, we're happy to look at whether they can be connected for you.

How far can one booking go?

Once a booking can trigger two steps, it can trigger more. Here are a few examples of what you could add to the same booking, with no AI involved at all:

• A text message to your phone the moment someone books, for when you're away from your email.
• A new customer folder created automatically, named after the person or business, ready for notes and documents.
• A row in a spreadsheet, so you have a running list of every enquiry: who, when, and what they wanted to talk about.
• A reminder to the customer the day before, so fewer calls are missed.
• Relevant files emailed to you before the call, such as a price list or the notes from last time, picked out by rules you set in advance.

Each one is just another program's counter being asked to do one more job. The important part is that you decide the steps once, and after that they happen every time without anyone remembering to do them.

What is MCP?

MCP stands for Model Context Protocol. It's an open standard, meaning an agreed set of rules that anyone can use, for connecting AI to the software you already use. The project's own introduction compares it to a USB-C port for AI applications: one agreed shape of plug, instead of a different cable for every device.

Here's how it works. Every API is a slightly different counter, with its own way of asking for things. MCP puts a standard front on each one, a bit like a receptionist who knows that counter well. The receptionist tells the AI what it's allowed to do there: "you can read the calendar", "you can search these folders", "you can draft an email". The AI can then use any of those abilities when it needs them, and only those.

The difference from an ordinary automation is who decides the next step. The steps in our booking setup were written in advance and never change. With MCP, the AI can look at what's in front of it and choose which of its allowed abilities to use. We've written more about where that line falls in what an AI agent is.

What AI could add to the same booking

This is where it gets genuinely useful, because some jobs around a booking need a bit of judgement that fixed rules can't give. With AI connected through MCP, the same booking could lead to things like:

• A short brief before the call. The AI reads what the person wrote when they booked, looks up their business's public website, and checks whether they've emailed you before. An hour before the call you get one page covering who they are, what they asked about and anything worth knowing.
• The right files, found by meaning, not by rules. Rather than attaching a fixed price list, the AI reads the booking and finds the documents that actually relate to what the person asked about, then sends them to you with the brief.
• A reminder that notices problems. If a booking lands on a day your calendar shows you're away, or clashes with something added later, the AI can spot it and draft a polite message offering another time, ready for you to check and send.
• A question sent ahead when the booking is vague. If someone just writes "AI stuff" in the booking box, the AI can draft a friendly note asking what they'd most like to cover, so the call starts in the right place.
• A follow-up drafted after the call. If the call is recorded with the other person's agreement, the AI can turn the notes into a summary and a draft follow-up email for you to read before anything goes out.

Notice that in most of those, the AI drafts and you decide. That's deliberate. MCP lets you choose exactly what the AI can touch, so it can be allowed to read your files and calendar while anything that goes to a customer still waits for a person to approve it. Giving it only the access it needs is the sensible way to start.

If you'd like something like this set up around your own bookings or enquiries, we're happy to build it and set sensible limits on what the AI can do.

Where to start

Start with the plain automation. Get the steps you already do by hand, like the email, the calendar entry and the spreadsheet row, happening on their own and working reliably. Then add AI where a job needs judgement rather than rules, one piece at a time.

You don't need to understand APIs or MCP to benefit from them, any more than you need to understand wiring to use a plug socket. It's enough to know that if your software has an API, it can almost certainly be connected, and that AI can now join those connections in a controlled way. If you'd like to see what that could look like for your business, have a look at what we do and book a free call. Your booking will set off the same two steps described at the top of this page.

Sources

• Introduction to the Model Context Protocol - Model Context Protocol

Checked on 22 September 2026.