If you've been anywhere near a tech news feed lately, you'll have seen the letters MCP popping up with increasing frequency. Model Context Protocol. It sounds like one of those things that's been specifically designed to make your eyes glaze over, and we wouldn't blame you for scrolling past it.
But here's the thing. MCP is one of those rare bits of technology that actually matters to businesses, not just to developers. It changes the way your AI tools connect to the systems you already use, and it does it in a way that's worth understanding, even if you never plan to write a line of code in your life.
We're using MCP servers in anger, and we're blown away by their simplicity and functionality. We think you'll love them too. So let's do what we always do and explain this properly, in plain English.
So what actually is MCP?
MCP stands for Model Context Protocol. It was created by Anthropic (the company behind the AI assistant Claude) and released as an open standard in late 2024. Put simply, it's a set of rules that lets AI tools have proper conversations with your business systems.
You've probably heard people call it "USB-C for AI", and that's not a bad way to think about it. Before USB-C, you had a different cable for every device. Your phone charger didn't fit your laptop, your camera used something else entirely, and you ended up with a drawer full of cables that were all slightly different. USB-C fixed that by creating one standard connection that works for everything - note to self, update iPhone.
MCP does something similar, but for the way AI tools connect to your software. Instead of building a bespoke connection between your AI assistant and your CRM, then another one for your accounting package, and another for your project management tool, MCP gives them all a common language. One standard way to talk to each other.
Why "conversations" and not just "connections"?
This is the bit that actually matters, and it's where MCP is properly different from what came before.
You're probably familiar with APIs, or Application Programming Interfaces, even if only vaguely. An API is basically a way for two pieces of software to swap data. Your website communicates with your CRM via an API. Your email marketing platform pulls contact lists through an API. They've been around for years, and they work fine for what they do.
But APIs are a bit like sending a letter. You write a very specific request, post it off, and get a very specific response back. If you want something slightly different, you need to write another letter. There's no real back-and-forth, no understanding of context, no memory of what you asked five minutes ago.
MCP works more like an actual conversation. When an AI tool connects to one of your systems through MCP, it doesn't just get a list of data. It discovers what it can do, what information is available, and how to ask for what it needs. The system essentially introduces itself: "Here's what I am, here's what I can do for you, and here's how to ask me." And the AI understands all of that in real time.
That's a meaningful difference. It means your AI assistant can do things like check your CRM, look at a customer's history, draft a response based on that history, and update the record afterwards, all in one fluid exchange. Not because someone painstakingly coded every step, but because the AI and the CRM are having a genuine back-and-forth about what needs to happen.
What does this look like in practice?
Let's say you're a managing director at an engineering firm. You've got a CRM (maybe HubSpot, which we've written about before), a project management system, and an accounting package. Today, these are basically separate islands. Your team copies information between them manually, or you've paid for some integrations that move data around on a schedule.
With MCP, an AI assistant could sit across all of those systems at once. You could ask it something like "Show me all the projects for clients who have outstanding invoices over 60 days", and it would know how to check your project management tool, cross-reference with your accounting system, and give you an answer. Not because someone built that exact report, but because the AI can talk to both systems and figure out how to get what you need.
Or think about something simpler. Your marketing manager wants to know which blog posts generated the most leads last quarter. Right now, that probably means pulling data from your website analytics, cross-referencing with your CRM, and spending half an afternoon in a spreadsheet. With MCP-connected systems, an AI assistant could have that conversation across your tools and give you the answer in seconds, pretty much.
If you're already thinking about how AI fits into your marketing, our piece on AI search strategy for marketers covers some of the broader picture.
How we're using MCP Servers
We're not just writing about MCP from the sidelines here. We've been using it to connect our own systems, and it's a good example of what this looks like in practice for a business our size.
Like many agencies, we've got several systems that need to talk to each other. We built our own business management platform, REDMan, where we track client work, budgets, and deliverables. We use Monday.com for day-to-day task management across the team, and HubSpot as our CRM for managing client relationships and marketing activities. Three systems, three different jobs, and, until recently, essentially three separate islands of information.
Using MCP, we've connected all three. Our AI assistant can now pull client work data from REDMan, create the corresponding tasks and groups in Monday.com, and cross-reference everything with the client record in HubSpot. What used to mean someone manually copying deliverables from one system into another (and inevitably getting something out of sync) now happens through a single conversation between the AI and all three platforms.
The bit that surprised us was how much faster it came together compared to traditional integrations. We didn't need to build a custom API connector for each pair of systems. MCP gave the AI a way to understand what each system could do and work out how to move information between them. It's not magic, and it did take some setting up, but the difference in time and effort compared to building bespoke integrations was significant.
We're sharing this because we think it's useful to hear from someone who's actually done it, rather than just read another theoretical article about what MCP could do someday. It's working for us now in production with real client data. It's not perfect (the security and authentication side still needs work for some use cases), but it's already saving us hours every week.
Why should you care right now?
We'll be honest, MCP is still in its relatively early days. It was only released in late 2024, and while adoption has been fast (Google, Microsoft, Salesforce, and most of the major AI providers have backed it), it's not like you need to rush out and overhaul everything tomorrow.
But there are a few reasons it's worth paying attention to now rather than in two years' time.
First, the cost of connecting AI to your business systems is about to drop significantly. Early reports suggest companies are seeing integration timelines cut by a factor of ten. That's the difference between a six-month project and a few weeks of work. For mid-market businesses that don't have enterprise IT budgets, that changes the maths on what's worth doing with AI.
Second, it reduces your dependency on any single vendor. Because MCP is an open standard, you're not locked into one AI provider's way of doing things. If you build your connections using MCP, you can swap out the AI tool at the other end without rebuilding everything. That's the kind of flexibility that saves you money and headaches down the line.
And third (and this is the big one, in our opinion), it makes AI actually useful rather than just impressive. We've all sat through AI demos that look amazing in a sales presentation and then turn out to be fairly limited once you try to plug them into your actual workflow. The reason for that gap is almost always the same: the AI couldn't connect properly to the systems where your real data lives. MCP closes that gap. It gives AI tools the ability to work with your real information, in your real systems, in real time.
What about security?
This is always the question (and rightly so). If your AI assistant can talk to your CRM, your accounts system, and your project management tool, who's making sure it doesn't do something it shouldn't?
MCP was designed with this in mind. The protocol includes controls that let you decide exactly what an AI tool can and can't do. You can give it read access to your CRM, but not write access. You can let it see project timelines, but not financial data. Every action gets logged, so you've always got a clear audit trail of what was accessed, by which AI tool, and when.
It's not a free-for-all. Think of it more like giving a new employee specific permissions. They can access the files they need for their job, but they can't wander into the finance director's office and start rifling through the drawers.
That said, the enterprise security features are still maturing. Authentication, single sign-on integration, and governance tools are all on the 2026 roadmap. If you're in a regulated industry, it's worth keeping an eye on these developments before going all-in.
Where does this leave you?
If you're a business owner or MD trying to work out where AI fits into what you do, MCP is basically the infrastructure that's going to make AI practical rather than theoretical. It's the thing that turns AI from a clever chatbot into something that actually works with your business, day to day.
You don't need to understand the technical details. You don't need to know what a protocol specification looks like or how server handshakes work. What you do need to know is that the barrier between your AI tools and your business systems is getting much lower, much faster than most people expected.
The businesses that will benefit most are the ones that start thinking about this now. Not necessarily building anything yet, but getting their data in order, understanding what their systems can and can't do, and having conversations with their tech partners about what MCP-ready looks like. If you've been meaning to sort out your approach to AI and search, or even just get your website working harder as a lead generation tool, this is another good reason to start.
Frequently asked questions
What is MCP (Model Context Protocol)?
MCP stands for Model Context Protocol. It's an open standard created by Anthropic that gives AI tools a common language to connect to your business systems, such as CRMs, accounting packages, and project management tools. Instead of building a separate integration for each tool, MCP provides a single standard way for AI to talk to all of them.
What is the difference between an MCP and an API?
An API is like sending a letter: you make a specific request and get a specific response back, with no real back-and-forth. MCP works more like an actual conversation. The AI discovers what a system can do, what data is available, and how to ask for what it needs, all in real time. MCP doesn't replace APIs; it sits on top of them and makes them work properly with AI tools.
Is MCP secure for business use?
MCP includes built-in security controls that let you decide exactly what an AI tool can and can't access. You can grant read-only access to some systems, block access to financial data, and every action gets logged with a full audit trail. Enterprise features like single sign-on integration and governance tools are still maturing and are on the 2026 development roadmap.
Why does MCP matter for mid-market businesses?
MCP significantly reduces the cost and time needed to connect AI tools to your existing business systems, with early reports showing integration timelines cut by a factor of ten. Because it's an open standard, it also prevents you from being locked into a single AI vendor. For mid-market businesses without enterprise IT budgets, this changes the maths on what's worth doing with AI.
Do I need to do anything about MCP right now?
You don't need to rush out and overhaul anything tomorrow. MCP is still in its relatively early days, though adoption has been fast with backing from Google, Microsoft, and Salesforce. The best thing to do now is get your data in order, understand what your current systems can and can't do, and start talking to your tech partners about what MCP-ready looks like.
We're keeping a close eye on how MCP develops, particularly as it relates to marketing platforms like HubSpot and the tools our clients use every day. If you want to chat about what any of this means for your business specifically, get in touch. No hard sell, no jargon, just a conversation (a human one, not a protocol one) about where things are heading.
