194 | MCP Servers Demystified: The Simple Way to Supercharge AI for Your Business with Kate Shevchenko

0:00Hello, and welcome to another live episode of the Leveraging AI podcast, the podcast that shares practical ethical ways to leverage AI to improve efficiency, grow your business, grow your business, and advance your career. This is Isar Mektes, your host, and I've got an incredibly fascinating and exciting episode for you today. Those of you who've been following the podcast know that we've been talking a lot about MCPs. Now, those of you who are not familiar with what MCP is, what happened is if you think about the whole world of AI agents, it started with an idea that allows you to then connect different tools and systems into an agent that can help you work through different systems and processes in your company.

0:39The problem was that every single company, let's take Salesforce as an example. I don't know how many millions of companies in the world use Salesforce, but every single one of them, if they wanted to connect their Salesforce into their AI tools, whether it's a chatbot or an agent, needed to create their own integration. That's a pain in the behind, and it is very costly, especially if you add the amount of time that every single company has to do this. Well, a few months ago, earlier this year, Anthropic came up with a standard protocol on how to do this. And the idea is you can create one integration for Salesforce that is going to be an MCP server that basically on one end connects to Salesforce as a whole, any Salesforce.

1:20And on the other hand, you can connect it very easily in a matter of minutes, and if you're really good, seconds, into your AI universe. And now your AI universe can talk to your Salesforce without having a team of five developers work for six months in order to create this integration. This is magical. Now, what happened, because they made this open source, many, many companies on the planet started creating MCP servers for their tools, basically exposing their tools, their data, their information, their capabilities to your AI tools that you're using, whether it's ChatGPT, Claude, or an agent that you're building.

1:53And so this took the world by a storm. Now, to be fair, most common people, myself included, did not use MCPs at all, even though we heard the hype, but it sounds really complicated. How do I connect to this? And then I saw a post on LinkedIn by Kate Shevchenko. Kate is the COO of a company that's called Data for SEO. And they provide a really amazing and very cost-effective tool that actually competes with the big tools like SEMrush or Ahrefs, but with a much better structure. And they created their own MCP server. And her demo showed how to connect Claude to their MCP server in a minute. Like, okay, now I'm curious.

2:31I want to know how to do this. How can you take a whole universe of data knowledge and capabilities, in this particular case in SEO, and connect it to Claude that I can now query any SEO question that I want just in a Claude chat? And so I invited her over. And so what we're going to talk about today, we're going to talk about what is an MCP, what are the benefits of having an MCP server for your company and for your business. But we're going to dive in into how to actually connect an MCP server, in this particular case, their MCP server, to both Claude and ChatGPT, whichever one you're using, in just a matter of minutes.

3:06I can tell you that once I followed the process, after Kate showed me how simple it is to do, I started creating other MCP servers and connecting them, like Slack and different data sources that I have, and CRM and things like that. Like, oh my God, this is like an endless universe of capabilities. So I'm very, very excited to welcome Kate to the show. Kate, welcome to Leveraging AI. In the next few years, AI technology will change our world dramatically. Whether you are a business executive trying to catapult your business forward, or just somebody who refuses to be left behind and want to advance your career, this is the show for you.

3:46I'm your host, Isar Matis, a serial entrepreneur and an AI enthusiast. You'll hear invaluable practical tips from innovative business leaders, AI practitioners, and some of the brightest AI minds in our world today on how you can leverage AI in ethical ways to advance your career and grow your business.

4:09Thank you very much. Thank you for this introduction. And I agree on everything that you're saying about MCP, how huge it is right now. and I'm really excited to be here and share our experience of a company which created MCP and how it's going on right now. So spoiler alert, it's like a huge hype right now because you noticed me because of my post on LinkedIn and we didn't expect such a lot of people there. We knew that the MCP is a huge stain right now but we didn't expect such a lot of people as Beta for SEO is quite a technical product because we sell data through APIs. We are a B2B business mostly.

4:51That's why we don't have like a huge attention as we could because we only know among SEO tools and SEO agencies or marketing agencies. But such things like no-code integrations or MCP servers make us visible. We are pretty excited about it. When I started working in DataFresh, it was really difficult to get started because I'm not techie at all. But right now, I believe it's easier, much easier right now to get started with such things, with APIs, because you have a mega.com, NA10, you have MCP servers, which you connect to cloud or change the key and so on. So really cool thing. Yeah, I agree with you.

5:36So going back to what you just said and what I said before, both Kate and I are not technical people. Like we understand technology, but we're not developers. We don't have any engineering degrees. I've never written a line of code in my entire life. I don't know how to read code. I can use a lot of AI tools to do a lot of coding tasks, but I can't debug when it's getting wrong. I'm just telling it what went wrong and then it will go and fix it. But what you will see in today's episode is that it's actually very, very basic. Like even if you're afraid of the AI writing its own code, it's literally less than a minute and you can be connected to this amazing SEO tool.

6:12And like I said, with a similar effort connected to any tool that has MCP server today, which right now it's most major software companies already have MCP solutions that you can connect to and then bring them into your AI universe. Before we get started and dive into exactly how to do this and what are the benefits, I want to thank everybody who's joining us live, whether you are joining us on the Zoom call or whether you're joining us on LinkedIn. Please go introduce yourself right where you are, where are you joining from? What are your interests in this? Are you already connected to MCP servers?

6:44What are you trying to get out of this episode? Because that's going to help us a lot in kind of like pushing the call into the relevant direction. As for one last thing, as you all know, we're running the AI Business Transformation course. We just opened another cohort. So the May cohort is running right now. We do teach these courses all the time, at least once a month, but most of the courses are pre-sold to specific organizations and businesses. so the next public cohort is going to be in the second week of august and if you want to join the course then there's going to be a link in the chat and the link in the show notes if you listen to this after the fact every one of these courses is growing so this particular one has like 25 people i assume we're going to have about 30 and i don't think we're going to let more than 30 in because it's starting to get too many so if you want to join the course you might as well join early and make sure you have a seat but with that i'll give the stage and the microphone back to kate and kate Let's talk about the magic of MCPs.

7:37Let's start with really, what is it exactly? So people have a better understanding than my two seconds of introduction in the beginning. And then we're going to dive into how to actually connect to your MCP. Yeah, I can see some questions already. Yeah, MCP is the cool thing that makes life easier when it comes to connecting some data to AI agents. So we provide API solution. That's why I would stick to APIs more. mcp helps ai agent to understand apis better to process their data better and send a correct request to api of course you can be a developer and develop this connection by yourself you need to explain it to ai agent how api works which data can be sent there which response we can get and so It can be done for sure, but it would consume some time, much time, I would say.

8:35And it's not always that you send the right API request through the system that you create. Because the AI agent would send some random field, random parameters to API, which even do not exist because of a prompt which is sent by user to this AI agent. And that's why it's a really complicated thing. And if you need to connect API of one company and another and the third one, so it's much time you need to spend and every API is different, different structure and so on. It becomes a mess. If a company which provides APIs and they also have MCP supported, you just need to connect these MCPs. It would take maybe an hour for you to figure out how to connect everything and from different companies.

9:30But they have the similar structure. All of them have the similar protocol. MCP is a protocol. Your AI agent becomes smarter. It works better. It works faster. And it doesn't make mistakes because they know how to send the right request to these particular APIs. ads. The end user would see consistent and correct data. Awesome. Great explanation. Again, for those of you who are not technical at all and don't know what APIs are, so API is basically the language in which one software can talk to another software from a server-to-server perspective. And it's just a list of a very, it's a very long list usually of functions and how to call them and what parameters they need in order to actually work.

10:14And as Kate mentioned, yes, every API usually has a lot of documentation. You can go through documentation and write a lot of code to do this. But what MCP, the magic of MCP is it's basically a standard dictionary. And what that basically means is think about that you can speak in English and on the other side, you will speak in whatever language is required in order to make the communication with a specific software that you're talking to. And that's the beauty of this. And that's why the implementation became from, I need a team of three people to work for two weeks to create this connection to a non-technical person like me or like Kate can do this in a minute or two.

10:50And because everything behind the scenes is already predefined. So let's talk before we dive into the actual implementation and how to do this in Claude and how to do this in ChatGPT. Let's talk for a minute on your software, because I want people to understand what does MCP enable, right? So I think we can use your software as an example, and then I can potentially generalize it to other capabilities. But let's talk about what your tool provides, regardless of MCP, and then why does MCP make it a lot easier to access? Yeah, we provide SEO and marketing data. For example, we provide ranking data, keyword research, keyword overview, book links, overview, domain overviews, and so many stuff which you can see in every SEO tool.

11:38I bet that once in a life you used some of SEO tools or marketing tools. So such companies as I, as ours, are behind such tools. So we work with SEO tools and we collect data for them. They connect to us through APIs. So they send requests to us, for example, you're in a tool typing, I want to get keyword suggestions for my keyword for James, I don't know. So the SEO tool sends a live request to us, to data for SEO. We go to our huge database of keywords. We process this request and we send the requested data in JSON format, keywords, and each keyword put search volume, CPC, keyword difficulty, and many, many other data.

12:25So that's it. So again, just to explain to people, this is not a tool you probably used, but you probably used other tools if you've done any SEO work that actually uses their data behind the scenes. And now, because of MCP, any person can now connect to that data. Any person can connect, yes, without any technical knowledge. Now they don't need to create a tool and connect its API to it. So now you can do some simple tasks in such tools like Cloud, ChatGPT, and so on. You can also connect it to NA10, for example, and develop your AI agent greatly and maybe send data to your Airtable, for example, Google Studio and many other things that you can imagine.

13:14Yes, you can be very creative with it because they're great for sure. They're sharing much data. They give you hints, suggestions of what to do for you to be better for Google to direct for your keywords. But sometimes they don't have functionality which you're looking for specifically for your specific use case. That's why such people go to API providers like us. They can build anything that they want. And right now, you don't need even to be techie to do it. You can go to no-code tool or MCP servers. Amazing. I'll generalize this for one second, and then we're going to dive in, and you're going to show us how to actually do this in Cloud, and we're going to do the same thing in ChatGPT, so you can prepare whatever you want to prepare on your site to demo.

14:00But to generalize this, what MCPs enable you to do is to get access to the backend, the data, the knowledge, the capabilities that software companies provide and query them in your universe, connected to your tools, your data, your needs, and so on. So what Kate mentioned, tools like N10 or Make and so on, it means you can, without writing a line of code, create a tool that will, now let's generalize it and not use SEO and grab data from your CRM, from your SharePoint and from your Slack channel because you're going to connect it to three different MCP servers and then ask any question you want about any project and it will give you the information combined from these three sources because it knows how to query each and every one of these sources and each and every one of these integrations is going to take you a few minutes to create.

14:53If you're really good, probably is less than that. So it's capabilities we didn't have before. Again, unless you had a huge development team And now literally every individual, forget about a company, can now do this without a crazy amount of effort. Definitely an effort that justifies the benefit that you're going to get out of that. So with that in mind, Kate, let's show people how to actually connect, how simple it is to actually connect an MCP server to Claude. And then we're going to do ChatGPT as well. For those of you, by the way, who are driving or on the treadmill or walking your dog because you're listening to the podcast and not watching this live, first of all, join us live.

15:27We do this every week at noon and you'll be able to actually see and ask questions that you have. By the way, just to tell you how crazy it is right now, we have people from Australia, which is really late, I think, at night right now in Australia. Kudos for Catherine from joining us from Australia, Adam from New York and people from all over the world are here. So you can come and join us every Thursday and then you'll be able to ask your own questions. But we're also going to explain everything that's on the screen so you can understand what we're doing. I can imagine because it's England for me here in Ukraine.

15:54So it's crazy with sign Australia. So thank you for joining everybody. Yes, for starters, you would need to download the cloud desktop. And also you would need a paid plan because they don't allow anything to configure. Additionally, if you have a free plan, it's obvious, but I need to mention this. Also, you need to install Node.js. You just need to install it and that's it. Nothing else should be done. You don't need to call Terminal, for example. You just need to install it and that's it. And by the way, it's as simple as going to the website, cleaning and download for your relevant operating system, whether you're Mac on this, cleaning, install, and you're done.

16:38Like there's no technical work to do there. And it's free. Like it doesn't cost anything. Yeah. So sharing a link to the chat, you can download and try in the barrel as I'm doing it. Yeah, we have everything that we need and we can start connecting MCP to cloud already. Isar, can you see my cloud? Right? Everything is cool. Yeah. Cool, cool. All right. We need to go to settings of cloud. Oh my God, my computer is tired in the evening. And you need to go to developer and you need to click on this button, edit config. So for those of you not seeing, when you're going to the setting in Cloud, you have two options.

17:25One is like the basic ones, and then there's developer ones. And then in there, there's just, if you click on the developer, there's only one button to edit the config file, which literally just takes you to your internal browser to find the config file. And then you can edit it with whatever text tool you have installed on your computer, whether it's Mac or Microsoft, literally any text note editor will do the trick. Yeah, we opened this config and we just need to copy paste one tricky config that will do all the work. So yeah, I sent it to you with email. Yeah, and we will share that. We need to edit this config with this one.

18:05So here you can see Data4ACO username and Data4ACO password. I already paste users' username and password, but you can get it after register at Data4Asia. You get free funds when you register. You don't need to worry about it, so you can test it for free. You can use also this credential as well. They are free. They have, like, I added some funds to it. So feel free to use it. And I will share it in the chat as well. So you just can copy and paste it. and we just need to save this thing. We go back to Claude and we need to quit it and we need to open it again. So before we jump back to Claude, let me mention what we actually did.

18:56So we opened the config file, we copied and pasted code that we don't know what it does. And the only thing we had to add in there is in the place where it asks for the username and the key, basically the password, We add in the username and the password that you're going to get if you sign up to their tool. It's a very similar process when you connect to any other MCP. Like I said, already connected to Slack. So in Slack, you create a new tool in Slack and it's going to give you your username and it's going to give you your unique key to use. So these are the only parameters you need to change.

19:26Otherwise, the code is exactly the same piece of config, which is like five lines. Like it's nothing crazy that you can copy and paste and just add in there. And then you'll be able to connect to multiple different tools. Yeah, right. So if you reopen your cloud, here you can see that data for SEO is installed already. So you have like 47 endpoints connected right away in just one minute. and for again just just one thing to understand what end points are so in claude you have these little toggle buttons icon within the chrome within the chat box if you click on that it shows you what you can connect to so are you connected to web search are you connected to google drive and underneath that you just add another other option which is data for seo and what end points are is basically functions so there's 47 you said 47 different functions that you can call, whether it's to find keywords in a specific page, find keywords in another people's page, do an SEO analysis, who are the top ranking websites for whatever, like 47 different SEO related questions.

20:28And it's a similar thing with other tools. It's just going to bring different functions that you can call. Now, you don't need to know the names for these functions because it's an AI chat. You can just ask in simple English. So let's do a quick demo of what you can do now that we are connected. yeah for example you can get a list of keywords for which your website ranks for you can get any number for example we can use just our website like return top 20 runkit keywords for data for asia.com

21:08simple prompt in English and that's going to call the MCP server because it understands that now it has that tool connected to it which is brilliant right we don't need to actually tell it to go and use that tool just because it's connected it understands that that's where it needs to go yeah it's why Claude is cool is that you can see which function is calling a data for SEO and you can see its request, actual request that it makes and the response as well. So it's a technical thing which regular SEO tool would use when they want to connect APIs. But as a regular user, you don't need it. So here that is, you can see top 20 ranked givers for my website.

21:49For each one, I get position and the page. Also get search volume difficulty and also I get traffic value for my page. which is bank for. So yeah, that's it. We can also get... Just to explain what we just did, in copying and pasting one piece of code that is like less than 10 rows long into the config file of Claude, which again, it sounds technical, but it's not. It's literally copying and pasting and getting the username and password from Data4SEO on their website as you register. you can now ask any SEO question about any website or webpage and get actual data that a lot of other companies are using in the world straight into Claude while asking simple questions in English.

22:40And as I mentioned, if you broaden that, if you look at this from the 30 ,000 foot view, the same thing is for any tool you use out there, as long as it has MCP support. If it doesn't, it probably will sometime in the next few months because everybody will. There's a question in the chat is this available on the free version of cloud and the answer is no you need a paid version of cloud in order to do this i believe that you can create custom gpts in the free version of chat gpt so maybe that's a workaround if you don't want to be paying for any of these tools yeah but our investing in cloud is a great thing to do as it's just a simple prompt that you can see, but you can ask to analyze your website with our APIs and give you a strategy for your SEO work.

23:28So many, many possibilities which space plan of cloud can do, which is not possible on the free version. Yeah. And I just did this, by the way, for my website. And it's incredible the amount of insights and the level of details it go to from which H tags exist and don't exist on specific pages, which tags I should remove or add, depending if they're duplicate or they're missing. How am I doing on the percentage of text versus the overall code of the page? Like a lots and lots of really useful SEO stuff that most people, unless you're an SEO person, doesn't even know to look for. And with one simple request and dropping in my link, I got a very detailed analysis that I would have paid a lot of money to a company to do for me.

24:10And so it's really, really amazing. uh yeah and you wanted to see it in action in change gpt as well right yeah so let's do the chat gpt integration over there it's even easier so if this wasn't easy enough and maybe you don't have an integration to maybe you don't have a paid version of claude then you can do this in chat gpt as well so really let's dive into chat gpt so again for those of you who are listening we're now opening chat gpt and i will let kate walk us from here oh cool so yeah we need to go to check GPT and if I'm not mistaken, it also requires paid plan to do these GPTs. But you need to check for sure.

24:49We go to the GPTs and we need to create a new one. We go here. We just name it as we want. Demo, for example. And we create a new action. We need to specify our API login and password, which we have a data for ASIO. So we choose API key here and specify our password and login and password here. Yeah, I'm doing it right now. Serving in the chat as well. You can use this one. It would work. Then we go to this URL. We have this GitHub page serving in the chat as well. You just need to copy these actions and paste here. And we already can see some available actions. We just hit test to see if it's okay, if our credentials are fine.

25:49We click confirm. Yeah, we can see that it's working. Yeah, the default data is shown. We click create only for me and it's ready. So we can go and ask some questions with the chat GPT to get some SEO insights. So again, to explain the slightly different approach here, instead of actually creating an MCP connection, we are creating an MCP connection, but we're creating it as part of a portion of a custom GPT that most people don't use. So those of you who don't know what custom GPTs are, you should. These are the best magic that was provided to business people ever, probably. But it's a tool within ChatGPT that allows you to create these custom automations that can run again and again and again.

26:35and one of the functions, and you just give it a name and you give it instructions in English, but there's a bottom section to all of those that allows you to connect it to different API and external tools. And really what Kate showed us now is that there's a standard code, again, don't need to understand, that you can copy and paste into that segment to basically create a custom GPT that in the backend connect to their MCP server and can do everything that we showed on Cloud before. It just runs as a custom GPT within ChatGPT. I did check and you are right and I was wrong. You cannot create custom GPTs with a free license.

27:11That being said, you can use custom GPTs with a free license. So if you have one person in your company that has a paid license, they can create it for you and then you'll be able to use them moving forward. Thank you. And I also received a message from our developer who made this all in life. She told me to love for it. And he said that it's possible to use this in cloud in free version, but you need to specify and get less endpoints because it's limited to the number of tokens they can process. So you just need to do it from your end and just maybe collect two or three endpoints and you can go and test it.

27:52All right. So, yeah, I just made a random check of this page. for SEO on page health. It shows also the request and we can see some statistics on it, something that we need to improve and so on. So you can see it. So we can go further and communicate with chat. For example, we can ask to suggest which keywords we want to target for this page, which is the possibility to get into top 10 of Google, which is search volume and so on. Just to say how incredible this is, think about having a custom GPT or a conversation in Claude that will allow you to dump in your competitor's website, get exactly what they're ranking for, ask for suggestions on what keywords you should try to go for with long tail keywords, and then potentially get recommendations of what content to write in the same chat.