Build Your Own MCP Server with Databutton (Claude Desktop, Cursor, OpenAI)
Other Videos📅 2025-04-04
Transcript
[00:00] Let's learn in this very simple guide how we can start setting up MCP servers within our data button apps. If you're familiar with the channel, then you already know we've created software together using data button. So, I'm going to take the current software of Mimik here and add an MCP server to it. So, if that all sounds good, let's go and jump in. Welcome back, y'all. Today's video is sponsored by Data Button. They hit me up and they're like, Corbin, we got a big update. I'm like, what's the update? They said M to the C to the P. And I'm like, okay, MCP server, let's set it up. So, let's go and check it out. In addition, I'm going to leave this link in the description down below. This goes over a ton of cool
[00:30] capabilities we can do with data button and MCP servers. We got a little cool demo here with Viral, one of the co-founders. Hello, Viral. Here are some of the use cases. We can do sales lead outreach tool, influencers research tool, performance marketing tool, and a ton of other cool stuff. I'll also be leaving a link to this as well. This is like a nice little template that just gets you going so you don't have to do all the prerequisite stuff when creating a datab. Let's just use a template, see if we like it, and proceed. Time to make my software into an MCP server. Before I dive any deeper, you might be asking
[01:02] yourself, Corbin, what is an MCP server? Why do I keep seeing it? What's going on? I'm going to have help docs and articles in the description down below so you can really dive deeper into the topic of MCP server. But here is the gist. All you need to know is that when you create software, eg the software I'm about to show you now with mimic and using data button, we can access these value points of whatever it may be. So in this context we enter in a brand voice and it gives us different social media copy in external applications which basically means that when I put a
[01:34] Nike tweet here. So if I put a Nike here so for example I come over to Nike I grab this tweet I put it in my current software which purpose is to grab authentic tweets from businesses generate content based off that that I could use in my own business. That's why it's called mimic. And there we go a tweet that is similar and in the brand voice of Nike. Now this is all enclosed within my 73 application of mimic. What an MCP server allows us to do is create a layer that accesses this information in external applications like cloud desktop. Everyone you feed you see MCP
[02:06] MCP we're going to go ahead and follow the help article provided by data button. I'll leave this in the description down below. We can walk this line by line so if you get lost you can reference this as well. First things first is you might be like Corbin how the heck did you even get to the point of building out this software where it is now? I actually did a whole video dedicated to this. It's how to use AI agents to create software. It's 18 minutes long. I'll also leave this in the description down below. So, if you want to know how we got to this point, whole separate video. Check it out. Let's convert this into an MCP server. I realize I'm going to say MCP a lot. Not
[02:37] the ABCs, the MCPs. So, to do that, all we need to do, and if you already have an existing datab account, this is perfect because it's a very simple integration. So, we're going to go to settings here. We're going to come up to production. And here we go. Step number one is we're going to hit enable MCP. Perfect. This is going to create an API key that we can therefore access in other types of software. So let's get our API key. API key is basically you you're knocking on a door. You're like, can I come inside? If you don't have a key, it's not going to give you access to the data. So we need this key to get access to the data.
[03:08] And in this context, that's that mimic data. Mimic the software application. It goes without saying, don't publicly share your API key. This should be a secret key. Keep it a secret. Once our key is created here, we have the ability now to leverage this MCP server in a bunch of different context, whether that is Claude, Cursor, OpenAI agent or Agno SDK. In this video, I'm going to show you how to leverage this with cursor. So, coming over to cursor here, I'm going to simply go to the cursor settings. To do that, I'm going to come up here to cursor settings, cursor
[03:39] settings. Once we're here, we're going to select the tab of MCP. As you can see, I have actually already connected the data button app. But let's go and do that for you. All you're going to do is come up here to add new global MCP server. Now, I'm going to move this little chatbot all the way to the left here so you don't see my key for obvious reasons. But all you need to do is come over to claw desktop. Simply copy this code snippet here. Copy snippet. And then paste it right there. What's nice is that data button will automatically fill in the relevant API key. So that's all handled. Once that's done though,
[04:09] what you'll notice is that you will get, you know, a nice little UI box here that says your data button app. In addition, you'll get access to the specific tools found within your server. So, as we know with the app that we created together called mimic, we have the ability to generate content. And generate content was this whole situation, right? So, I go to Nike, you know, I put in like the reference content of a Nike tweet and then I get the generated content to mimic a Nike tweet. So the idea now is that this MCP server that we just created together, we can access that value point of generated content
[04:42] anywhere. So what's super cool is that now that this is connected, we can access our own custom software API that we created, like our own internal API, we can access this outside of just the mimic app itself. So let's do that. So coming back over to X here, we can scroll down and find another tweet. For example, we can try the tweet, you're already late to the future. Am I late? Is time just a social construct? I don't know. That's definitely above my pay grade. The idea though is that if you have a very in-depth datab, we're gonna
[05:12] have a ton of tools we can access in this context, right? Not just generate content. It could be like five, six, seven different endpoints. So, let's try this though. So, I'm going to go ahead and say, okay, let's mimic this tweet by Nike. You're already late to the future and use the tool generate content. Idea being is that we're going to send a payload to our internal software application through this MCP server and get any outputs and receivable of a mimicked tweet. Hit send. First thing it should do is identify that it's using the tool. So as you can see right here, it's calling the MCP tool of generate content. We can hit run tool and based
[05:43] off that reference tweet, we got the tweet here of tomorrow's victory is built in today's sweat. Nice. As I stated before, you can find the video showing how we created this software of AI agents in the description down below. This is pretty cool stuff here with this MCP server. This is pretty new and obviously you've probably been seeing a ton of it in your feed. Idea though is that typically in software when we create value in the context of for example generate content in that endpoint, it's usually private and internal within the software itself. eg
[06:14] we can only access that software's value through the actual software domain and you know user interface but now with MCP servers we're almost giving our ability for that API to be accessed through external softwares such as cloud desktop as well or cursor therefore the future is pretty cool so I'll see you in the next video data button MCP two random videos that's my face I'll see you in the next The