MCP In 26 Minutes (Model Context Protocol) — Transcript
Full transcript
- 0:00I learned MCP for you. In the past 68
- 0:02months, you've probably heard a lot of
- 0:04people talk about MCP. MCP is so
- 0:06amazing. MCP is such a game changer.
- 0:09Well, I must tell you, they are correct.
- 0:11I have taken a bunch of courses, build
- 0:14MCP servers and clients for our team and
- 0:16a couple B2B clients that we work with
- 0:18as well. We've also included a section
- 0:20on MCP in our AI agents boot camp. So,
- 0:22in this video, I'm going to explain to
- 0:24you the fundamentals of MCP, show you
- 0:26how to use MCP servers and how to build
- 0:28your own MCP servers as well, both with
- 0:30code and no code. Throughout this video,
- 0:32there will be little assessments, which
- 0:33if you can answer these questions, then
- 0:35congratulations, you'll be educated on
- 0:37MCP. Now, without further ado, let's go.
- 0:39A portion of this video is sponsored by
- 0:41Balt. Okay, a quick overview of the
- 0:43structure of today's video. First, we're
- 0:44going to define MCP, talk about why it's
- 0:46important, and how it is that we use
- 0:48them. Then we'll be diving under the
- 0:50hood in which I'll cover the
- 0:51fundamentals of MCP, the concepts and
- 0:53frameworks that govern it. Basically,
- 0:55everything you need to know to be able
- 0:56to use and build your own servers. But
- 0:58it is not enough just to know all these
- 0:59concepts. So I'll actually show you how
- 1:01to build one as well, both with code and
- 1:03no code. Let us now define MCP. MCP,
- 1:06which is the abbreviation of model
- 1:08context protocol, is defined by
- 1:10Enthropic, the people who made it, as an
- 1:12open protocol that standardizes how your
- 1:13LM applications connect to and work with
- 1:15your tools and data sources. Okay. What?
- 1:18So first time I read that definition I
- 1:20was like sounds important. So let me
- 1:23actually explain it to you the way that
- 1:24I understood it using an analogy. MCP is
- 1:27like this USB plug here. It allows any
- 1:30device to be able to connect to any
- 1:32other devices to be able to work
- 1:33together. Prior to this universal
- 1:35standard of the USB port, if you wanted
- 1:38to connect things to each other, like
- 1:39say your computer to your microphone, to
- 1:41your webcam, you would be having a very
- 1:43bad time because all of the ports to
- 1:45connect with each other would be
- 1:47different, which means you would have to
- 1:48get a bunch of different special cords
- 1:50and different adapters to manage to get
- 1:51them to work together. And then you
- 1:52would have to write the software to
- 1:54cater towards like those specific port
- 1:55thingies, etc. Very complex, very messy.
- 1:58But with this USB port, we're able to
- 2:00now just connect all these devices to
- 2:02each other and they just work. Amazing.
- 2:04Wow. Do not be fooled. It sounds like
- 2:06such a simple little thing, right? Just
- 2:07like standardizing the port connectors,
- 2:10but it amounted to massive amounts of
- 2:11innovation with this ease of use. Now
- 2:13before MCP existed, if you have your
- 2:16agent and you wanted to give it access
- 2:17to different tools and different context
- 2:19or different types of data, you would
- 2:20have actually faced a very similar
- 2:21problem to the pre- USB port days
- 2:24because each external system and tool
- 2:25that you're trying to connect to has
- 2:26their own specific definition of how it
- 2:28is that you interact with it, which is
- 2:30their special API. For example, if
- 2:32you're trying to make a scheduling
- 2:33agent, you would need to give it access
- 2:34to like your calendar, your email, a
- 2:37note-taking app, Zoom, Calendarly, and
- 2:39you would need to custom write the code
- 2:41for each specific tool that you want to
- 2:43give your agent, which is a pain in the
- 2:45ass. So that's why when Enthropic
- 2:46proposed the model context protocol, the
- 2:48MCP, which would standardize the ways
- 2:50that we interact with all of these
- 2:52external softwares, everybody was like,
- 2:54amazing. Wow, so good. Now your agent is
- 2:57able to have access to all of these
- 2:59different external softwares and tools
- 3:01using the same protocol. Within just
- 3:03months that the number of tools that you
- 3:04can give your AI agents now, which are
- 3:06called MCP servers massively exploded to
- 3:08thousands. Now there's over 20,000
- 3:10pre-built MCP servers that you can just
- 3:12grab and give your AI agent. Here is
- 3:14just a small sample. Now, anybody can
- 3:16easily write them, publish them, and use
- 3:18them. Let me actually show you exactly
- 3:20how easy it is to use an MCP server.
- 3:22Like say for example, I'm looking at
- 3:23these MCP servers here. You can pick any
- 3:25of them, but let's just pick like Alpha
- 3:27Vantage for example. It is a MCP server
- 3:29that enables LLMs and aentic workflows
- 3:31to seamlessly interact with realtime and
- 3:33historical stock market data through the
- 3:35model context protocol. Okay, great. So,
- 3:37you can get access to stock market data.
- 3:41You copy this stuff which allows your LM
- 3:43application to have access to the MCP
- 3:45server. copy it into your LM application
- 3:47like for example cloud code if you want
- 3:49to give the tool to cloud code and voila
- 3:51you'll be able to have access to the MCP
- 3:54server and all the different tools that
- 3:55are defined on the server then you can
- 3:57do something like plot the coffee stock
- 4:01market prices for the past 10 years then
- 4:05Claude will say Claude wants to use the
- 4:07coffee function from Alpha Vantage are
- 4:09we going to allow this and we'll be like
- 4:10okay allow once using that tool was able
- 4:13to get the coffee price data and now
- 4:14it's going to created interactive
- 4:16visualization and there you go, coffee
- 4:17global prices with all of this
- 4:19information. Oh my god, coffee is so
- 4:21expensive these days. So yes, this is
- 4:23pretty much how easy it is to use any
- 4:25MCP server now. And I can also very
- 4:27easily switch what is called the host
- 4:29which is cloud code in this case to
- 4:30something else like NA10 for example or
- 4:32my own custom AI app. So let's now
- 4:34actually dive into the foundations of
- 4:36MCP so you will exactly understand what
- 4:38just happened here and how to build your
- 4:40own MCP servers as well. And by the way,
- 4:42MCP servers are actually even more
- 4:43powerful than just giving your agents
- 4:45tools. There's a lot more that you can
- 4:47actually do, which is why people are
- 4:48able to actually build these MCP servers
- 4:50and then sell them.
- 4:53The best course that I found which
- 4:54covers the foundations of MCP is
- 4:56Anthropic's own course in collaboration
- 4:58with deep learning.AI called MCP, build
- 5:01rich context AI apps with anthropic. The
- 5:03course starts with explaining basically
- 5:05why is MCP important. So I already
- 5:07covered most of that. Um but I do like
- 5:09the diagrams that they gave which shows
- 5:10that see before MCP there was fragmented
- 5:13AI development with each AI app you
- 5:16would need to have a custom
- 5:17implementation custom prompt logic
- 5:19custom tool calls and custom data
- 5:21access. And you got to do that for all
- 5:22of them versus with MCP which is a
- 5:24standardized AI development for each MCP
- 5:27compatible application you're able to
- 5:29have access to a lot of different types
- 5:31of MCP servers. For example, you can
- 5:33have access to a data store MCP server
- 5:35so that you can have access to data
- 5:36stores. You can have a CRM MCP server so
- 5:38that you can have access to CRM systems
- 5:41and a version control MCP server so you
- 5:42can have version control software. All
- 5:44plug-andplay and the other way around as
- 5:46well. Every time you build an MCP
- 5:47server, it would be reusable by various
- 5:49AI applications like a Google Drive MCP
- 5:52server that has access to Google Drive
- 5:53functions that can be used by MCP
- 5:56compatible AI assistant, MCP compatible
- 5:58AI agent, and an MCP compatible desktop
- 6:00app or IDE too. All right. Next
- 6:02important concept to understand is the
- 6:03client server architecture. Within the
- 6:06components that make up MCP, there are
- 6:09three different components. The host,
- 6:11client, and server, HCS for short. And
- 6:14the way that they relate to each other
- 6:15is the host houses the MCP client, which
- 6:18goes through the MCP protocol in order
- 6:20to have access to the MCP server. Hosts
- 6:23are LM applications that want access to
- 6:25certain data or tools through MCP. As I
- 6:28showed earlier, the cloud desktop is an
- 6:29example of a host. NA10 is also an
- 6:31example of a host. You can also have
- 6:33IDs, AI agents. Really any large
- 6:35language model application can be a
- 6:36host. Now the host wants to use the MCP
- 6:39servers which are the lightweight
- 6:40programs that expose different
- 6:41capabilities like the servers here like
- 6:43alpha advantage we talked about earlier
- 6:45that's able to get real-time historical
- 6:47stock data. Uh there's also time
- 6:49provides time and time zone conversion
- 6:51capabilities. Cherry studio that's used
- 6:54for LM provider support like AI
- 6:56assistance and conversions, documented
- 6:57data processing. Postgress SQL to
- 6:59provide readonly access to Postgress SQL
- 7:01databases allowing schema inspection and
- 7:03execution of readonly queries, GitLab,
- 7:05Reddist, etc., etc. There is a lot. Now,
- 7:08in order for the host to actually have
- 7:11access to these MCP servers, they need
- 7:13to maintain what is called an MCP
- 7:15client, which lives inside a host and
- 7:17evokes the MCP protocol to maintain that
- 7:20one-on-one connection with the server.
- 7:22Got it? MCP client lives in MCP host to
- 7:25invoke the MCP protocol in order to have
- 7:28access to the MCP servers which the
- 7:30players being HCS host client and
- 7:33server. All right, time for a little
- 7:34quiz for the example that I showed
- 7:36earlier which allowed cloud desktop able
- 7:38to access historical stock data alpha
- 7:41vantage and cloud code. Which is the
- 7:43host and which is the server? Put your
- 7:45answer into the comments. Okay, let's
- 7:48now talk about what are the things that
- 7:49are actually contained within an MCP
- 7:51server. Well, there are three major
- 7:54things. These are tools, resources, and
- 7:56prompt templates. TRP tools, which are
- 7:59the things that we've kind of seen the
- 8:02most of are functions and tools that can
- 8:04be invoked by the client. So, these are
- 8:06these things over here like time series,
- 8:09intraday, time series, daily, blah blah
- 8:11blah, symbol search, global quote. Also,
- 8:13stuff like sending a message in Gmail,
- 8:15calculator, retrieval and search,
- 8:17sending a message, updating database
- 8:18records. All of these are tools. But
- 8:21that is actually not all. You can also
- 8:23have what are called resources which are
- 8:25readonly data that is exposed by the
- 8:27server. So this is the data that the
- 8:29client is able to query but cannot
- 8:31change. For example, there are certain
- 8:32files like markdown notes that's
- 8:34contained within the server that you
- 8:36might want to read. Maybe it's tracking
- 8:38logs. Say for example, every time you're
- 8:39calling a weather app tool and you're
- 8:41getting that data, you might want to
- 8:43actually save a record of all the
- 8:45different weathers that you've
- 8:46collected. So at some point you might
- 8:47want to compile that together and make
- 8:49like a visualization or something,
- 8:50right? It would not be very efficient if
- 8:51every single time you had to go collect
- 8:53a weather data point. You would have to
- 8:55like use the tool and spam the tool over
- 8:57and over and over again. It could also
- 8:58contain just database records of stuff.
- 9:00Like for example, there's just like
- 9:02information that you're storing like
- 9:03contracts, meeting recordings, and
- 9:05notes. Then there's also prompt
- 9:07templates. These are structured prompt
- 9:09blueprints so that you don't have to
- 9:10force your user to have to come up with
- 9:12their own prompts. Like for example, if
- 9:14you are someone who's trying to use an
- 9:15MCP server that uh is able to summarize
- 9:18like a certain transcript from a meeting
- 9:20and then generate a report based upon
- 9:22it, you could just write something like,
- 9:23oh, summarize this transcript into key
- 9:25action items, then generate a report,
- 9:28right? Like you could write something
- 9:29like that, but that probably won't yield
- 9:31the best results. So instead, the people
- 9:32that actually built the MCP server can
- 9:35put within their MCP server a prompt
- 9:37template that is much more flushed out
- 9:39that's able to actually extract the
- 9:41insights and generate the report in a
- 9:42much better fashion so that the user all
- 9:44they have to do is input like some
- 9:46specifics into that template. This takes
- 9:48the burden of the prompt engineering
- 9:50from the user. As a full example, you
- 9:52can have like a SQLite MCP server which
- 9:55gives access to like a database which is
- 9:57the SQLite database. Don't worry if you
- 9:59don't know what that is. is basically
- 10:00just like a database that can store
- 10:01information. Anyways, so you can have
- 10:03tools that allow you to read the SQL
- 10:06database to insert information into the
- 10:08database to change the information in
- 10:10the database and to delete information
- 10:11from the database. In addition to that,
- 10:13there is a resource that is logging all
- 10:16of the changes that are happening to the
- 10:17database so you have a historical track
- 10:19record of what's happened. And there's
- 10:20also some prompt templates that
- 10:22incorporates the best practices to
- 10:24interact with that database. Make sense?
- 10:25All right, time for our next little
- 10:27assessment. Please answer the questions
- 10:28on screen now and put it into the
- 10:31comments so you are following along. MCP
- 10:33is all about giving your agents more
- 10:35context so that they can do really
- 10:36powerful things. So when you're ready to
- 10:38deploy your AI agents with a lot more
- 10:40functionalities and you want really good
- 10:41UI, NA10 plus Bolt is a really powerful
- 10:45combination and you don't need to write
- 10:46a single line of code or you can just
- 10:48directly build via Bolt. Bolt can take
- 10:50you from an idea to a fully functional
- 10:52product within minutes. Whether that is
- 10:54software for your team, an internal tool
- 10:56or a website. It has everything that you
- 10:58need, whether that's hosting, backend,
- 11:00database, off, domains, and even SEO.
- 11:03One of the things I really like about
- 11:04Bald is that it seamlessly handles the
- 11:06entire process so that you can focus on
- 11:08building real things as opposed to
- 11:11handling infrastructure. You don't need
- 11:12to handle like setup, configs, war,
- 11:14external tools. Bolt puts the most
- 11:16powerful tools used by developers into
- 11:18your hands with an intuitive chat
- 11:20interface. It really is as simple as
- 11:22this. So, here we have the Bolt new
- 11:24landing page. And we can just type
- 11:26something like build an app that blurs
- 11:29out people's faces in pictures and just
- 11:33click build now and then let it do its
- 11:36thing. Of course, you can go into a lot
- 11:38more details and have a lot of like
- 11:39specific specs within the app as well.
- 11:42And look at that. Here we have this
- 11:44application and let's actually try it
- 11:45out. We will put an image. Here is a
- 11:49picture of me when I was still a little
- 11:50Tina. And there you go. We have the
- 11:52original and then we have the blurred
- 11:54version. And you can download the
- 11:56blurred image as well. As well as you
- 11:58keep building, you can include a lot
- 11:59more features and tweak things however
- 12:01it is that you want. Let Bald do the
- 12:03heavy lifting so you can focus on your
- 12:04vision instead of fighting bugs. You can
- 12:06try out Bald V2 for free today at this
- 12:09link over here, also linked in
- 12:10description. Now, back to the video.
- 12:12Okay, so the final concept that I want
- 12:14to cover for MCP before diving into
- 12:16actually building them is the
- 12:18communication life cycle. The way that
- 12:19the MCP client and the MCP server
- 12:22communicate with each other. The first
- 12:23step is the initialization when the MCP
- 12:25client interacts with the MCP server
- 12:27being like, hey, I would like things
- 12:29from you now and they initiate a
- 12:31connection. Then you have the message
- 12:33exchange phase when the MCP client makes
- 12:35a request to the server. for example
- 12:37like I want to use this tool and the MCP
- 12:40server goes like okay and gives it back
- 12:42which is the response. So you do that
- 12:44exchange the messages and then finally
- 12:46there's the termination phase when the
- 12:48client and the server terminate and stop
- 12:51interacting with each other. Now the way
- 12:52that the messages are done is called a
- 12:54transport. The transport handles the
- 12:56underlying mechanics of how messages are
- 12:58sent and received between the client and
- 13:00the server. And there are two different
- 13:01categories of how transports work. First
- 13:03one is when the server is running
- 13:05locally like your MCP server and your
- 13:07host all of this is within a single
- 13:09machine. The analogy that I like to give
- 13:11here is say that your MCP client is like
- 13:13the diner and the chef is like your MCP
- 13:16server who is going to be the one
- 13:18preparing the food for your diner. Now
- 13:19for the server and the client running
- 13:21locally, it'll be like just you and your
- 13:23friend cooking at home with like a
- 13:24notebook. You as the chef is just like
- 13:26cooking it and maybe you're just like
- 13:28writing down what it is that you're
- 13:29doing and like just passing a note to
- 13:30your friend the diner for some reason.
- 13:32So they're like keeping updated about
- 13:34what's happening while you're cooking.
- 13:35That is like the standard out a running
- 13:37log that's just like living locally
- 13:39between you guys. You don't need to rely
- 13:41on like a waiter or a cashier or
- 13:43anything like that. The other category
- 13:44of transports is for remote servers. So
- 13:46these are servers that live in other
- 13:48machines on the cloud somewhere else.
- 13:50And there are two different approaches
- 13:51to interact with a remote server. The
- 13:53first one is called HTTP plus SSE which
- 13:55is server sent events. Don't worry so
- 13:57much about the terminologies. It
- 13:58supports what is called a stateful
- 14:00connection. Going back to that analogy,
- 14:02in this case, you would have to like
- 14:03leave your house and say you're at a
- 14:04restaurant. You, as a diner, would sit
- 14:06down at the restaurant. You're the
- 14:08client, and you would interact with a
- 14:10waiter, which is a server, and you would
- 14:12tell it stuff like, "Oh, here's my
- 14:13table. Here's the appetizers that I want
- 14:15to order, blah, blah, blah." And then
- 14:16your server would like bring you those
- 14:17things, and say if you want to order
- 14:19like the same drink again, you could
- 14:20just say something like, "I'll like to
- 14:21have the same drink again." And you
- 14:22don't have to like exactly specify
- 14:24whatever it is that you ordered the
- 14:25first time. So this is called a stateful
- 14:27connection because the server maintains
- 14:28the state of it like remembering the
- 14:30knowledge the memory of what you ordered
- 14:31all this information throughout the
- 14:33entire interaction. Now there's also
- 14:35what is called stateless stateful
- 14:37stateless. This kind of transport
- 14:38supported by a streamable HTTP. To be
- 14:41stateless is sort of like if you go to a
- 14:43fast food restaurant as opposed to like
- 14:45a sitdown restaurant. Like if you're at
- 14:47McDonald's for example, you would order
- 14:48like a Big Mac and I don't know like
- 14:50fries. But then like if you want to get
- 14:52something else, you want to also get
- 14:53like chicken nuggets, you would actually
- 14:55have to go and like input your order all
- 14:57over again and get like a new ticket and
- 14:59you have to wait again. The cashier
- 15:00that's clashing you out, the server
- 15:01doesn't actually remember everything
- 15:03that you ordered previously. So that's
- 15:05why it's called stateless because it
- 15:06doesn't remember all of the knowledge
- 15:08that came with your prior interactions.
- 15:10Each request is independent of each
- 15:12other. So without going into way too
- 15:13much detail about this like the
- 15:14streamable HTTP is the preferred
- 15:17transport method because it's able to
- 15:19support both stateful and stateless and
- 15:21there are instances in which you want to
- 15:23have like stateful interactions and
- 15:25instances in which you want to have
- 15:26stateless interactions as well. So this
- 15:28section might sound a little bit like
- 15:29detailed and theoretical but it does
- 15:31become very important when we're
- 15:33actually going to be using our MCP
- 15:35servers and building our MCP servers
- 15:37which is right after I give you this
- 15:38final little assessment.
- 15:44Okay, I'm going to show you the no code
- 15:46implementation of building your very own
- 15:47MCP server now. And we're going to be
- 15:49doing this by using n10. So you can do
- 15:52stuff like send an email to Tina with
- 15:55the address tina@ lonelyotopus.com
- 15:59asking her if she wants to grab a coffee
- 16:04sometime.
- 16:07Execute workflow. Press enter. goes
- 16:11through this. We can see that it's
- 16:12accessing the AI agent is accessing the
- 16:14MCP client which is here. It contains
- 16:17the ability to have a calculator and
- 16:19also the tool that allows to send a
- 16:21message to Gmail. So, it's able to
- 16:23access the tool to send a message to
- 16:25Gmail so that you will be able to get
- 16:27this email. Hi Tina, I hope this message
- 16:30finds you well. I was wondering if you
- 16:31like to grab a coffee sometime. Let me
- 16:32know what works for you. Best regards.
- 16:33Email was automatically sent with NATO.
- 16:35Okay. So, how to build this from
- 16:37scratch? start off with a fresh new
- 16:39workflow using NA10. If you have no idea
- 16:41what NA10 is and you're not familiar
- 16:43with it at all, check out this video
- 16:44over here. I go into details about how
- 16:46to build AI agents using NA10. I'm not
- 16:48going to go into this here, so the video
- 16:49is not super long, but basically it's a
- 16:51no code low code tool that allows you to
- 16:53build workflows. Also allows you to
- 16:54build MCP servers. So you have your
- 16:56server trigger over here and you want to
- 16:58add some tools. Like for example, we can
- 17:00add the calculator tool, which is just
- 17:02like the ability to calculate things and
- 17:04do math. Not super innovative but yes.
- 17:07And then we can also add the email tool.
- 17:09So maybe like a Gmail tool. And in here
- 17:13we need to set up the credentials like
- 17:14create new credential. You can kind of
- 17:16go through the process. So all you have
- 17:19to do is like sign into your Gmail um
- 17:21which I will do here. And then you'll
- 17:23have the account that is connected. Yay.
- 17:26Then um because we just want to send
- 17:28emails here. So the operation here is
- 17:29going to be send. So for two we're going
- 17:31to let the model decide these. So, let
- 17:33the model decide the subject. Let's al
- 17:36also let the model decide uh we'll have
- 17:38the email type HTML and the message.
- 17:42Let's also let the model decide that as
- 17:43well. Wonderful. Amazing. So simple.
- 17:45Right now, let's save this. Now, in
- 17:48order to actually use our server, we
- 17:50want to come here, make sure that it's
- 17:52active. So, the server itself is going
- 17:54to be active. And we want to take this
- 17:57production URL. So, this allows access
- 17:59to the server. Now I have another
- 18:01workflow over here which is an AI agent
- 18:03in which you can prompt it directly and
- 18:05respond also attach the open AI chat
- 18:07model here. So in order to add the MCP I
- 18:10can click the MCP client tool the
- 18:12endpoint we're going to copy paste what
- 18:14we got from the MCP server and here in
- 18:17the server transport see we have the
- 18:19server send events which is deprecated
- 18:20and the HTTP streamable the two
- 18:22different transport right so we're going
- 18:24to click HTTP streamable transport and
- 18:26we don't need to change anything else
- 18:28over here. Wow. Now we have the MCP
- 18:30client. So let's actually test it out.
- 18:32Let's actually do math this time.
- 18:34Calculate 10
- 18:3711 * 99.
- 18:41We can see here that the AI agent goes
- 18:44over here. Human says calculate 11 * 99.
- 18:46It goes to the MCP client and it uses
- 18:48the tool cal the calculator tool which
- 18:50we have. And it says useful for getting
- 18:52the results of a math expression. And
- 18:54then the open AI chat model was able to
- 18:55get the results by using the MCP client.
- 18:58And then we were able to get the final
- 18:59result here with the output being the
- 19:01result of 11 multipi by 99 is 1,089.
- 19:05Yay. See how easy it is. There's a lot
- 19:08that you can do with this. Of course,
- 19:09you can add like so many more tools into
- 19:11your MCP server. And you also don't need
- 19:13to have NAN as your host. We can easily
- 19:16switch out the host for something else.
- 19:17Like for example, maybe you want to
- 19:19change out the host for cloud desktop.
- 19:20So what we need to do is go on cloud
- 19:23desktop. So go on help, type in
- 19:25developer. So switch to developer mode
- 19:28and then under developer open the claw
- 19:30desktop config file. So this should be
- 19:33empty. You can open it with whatever
- 19:35text editor. So I'm here I'm just using
- 19:36VS code for it. And you want to copy
- 19:38paste this. So this is defined MCP
- 19:40server and we're using NAN because
- 19:42that's where the server is and we are
- 19:44using SSE over here. And over here just
- 19:48paste the endpoint here. So command S or
- 19:52control S. Save it. Restart Claude and
- 19:56wow amazing. We have NA10 over here
- 19:58under the tools and includes send a
- 20:00message in Gmail and the calculator. And
- 20:02then you can do the same thing here.
- 20:04Send an email to Tina@ lonely
- 20:08octopus.com
- 20:09asking her to host an octopus party.
- 20:13Click enter.
- 20:15See it'll say over here Claude wants to
- 20:17use send a message in Gmail from NA10.
- 20:19And you can click allow once to allow it
- 20:21to use the tool from the MCP server. And
- 20:24moment of truth, we have over here, "Hi
- 20:27Tina, I hope this message finds you
- 20:28well. I want to reach out to see if
- 20:29you'd be interested in hosting an
- 20:30octopus party. I think it'll be a
- 20:32wonderful event and you'll be the
- 20:33perfect host for it. Let me know if
- 20:34you're available and interested. Looking
- 20:35forward to hearing from you. Best
- 20:36regards. Yes, automatically sent with
- 20:38NAT." Voila. See, super easy, right?
- 20:41Okay. So, now I want to show you the
- 20:42code implementation of doing this. Now
- 20:44with code you are able to do a lot more
- 20:47than this no code approach. I think the
- 20:49no code approach is amazing. It's great,
- 20:50right? It's really accessible and easy
- 20:52to use. Remember we talked about earlier
- 20:53about having like additional resources
- 20:55and also prompt templates. You as of
- 20:57right now cannot do that using NAT. You
- 21:00can't actually put in the prompt
- 21:01templates and specify those and you also
- 21:02can't put in additional resources as
- 21:04well. And in the code implementation you
- 21:06can. So I'm not going to go and show you
- 21:08every single line of code right now
- 21:10because I don't want to make this video
- 21:11like super duper long. But I want to
- 21:13show you like the snippet of these are
- 21:14the tools that we're going to be using
- 21:16and this is how you can actually get it
- 21:17to run. Okay. So for this code demo,
- 21:19we're going to be using cloud desktop as
- 21:22the host and I'm going to showcase a
- 21:24simple MCP server that's able to provide
- 21:26tools, resources, and prompts for
- 21:28working with Google Sheets and Google
- 21:30forms. So we can tell that the MC3
- 21:32server is connected because if you go to
- 21:34G Sheets over here, which is the MCP
- 21:36server name, you can see all the tools
- 21:37that are contained over here. So I have
- 21:39a lot of different spreadsheets that are
- 21:41available. So let's actually ask list
- 21:42out the spreadsheets that are available.
- 21:49It will want to use the list
- 21:50spreadsheets. So this is the tool and it
- 21:53tells us that we have 20 spreadsheets
- 21:55that are available in the Google account
- 21:57like NA10 form, anonymous feedback, the
- 22:00only octis alumni blah blah blah etc.
- 22:02All of these different ones. Would you
- 22:04like me to read the data from any of
- 22:05these spreadsheets here? I can say use
- 22:08read sheet to get the columns
- 22:13of the anonymous
- 22:17feedback spreadsheet.
- 22:20So it wants to use read sheet and this
- 22:22is actually a resource. So it allows you
- 22:26it's a readonly file where you can
- 22:28actually read what the column names are.
- 22:30You can see the spreadsheet ID and the
- 22:33information that's contained over there.
- 22:35And finally, I want to show you what it
- 22:36looks like to use some of the prompt
- 22:38templates. So, if we click over here,
- 22:40add from sheets, here are some of the
- 22:43prompt templates. So, let's click
- 22:44analyze sheet data. And we can say use
- 22:46this to analyze the anonymous
- 22:52pretty spell anonymous
- 22:55feedback
- 22:57spreadsheet
- 22:59and create a dashboard just to make it
- 23:02look pretty.
- 23:04And we can see here is what the prompt
- 23:06template actually looks like. And it
- 23:09shall do this. Let it do its thing. So
- 23:12while it's doing that, let me actually
- 23:13show you what the code looks like and
- 23:14how to run this MCP server. Here's the
- 23:16readme. If you follow these
- 23:17instructions, you'll be able to run this
- 23:18MCP server yourself. You can see that
- 23:21the tools that it contains include
- 23:23things like list spreadsheet, read the
- 23:25sheet, write the sheet, append the
- 23:26sheet, etc. It also has the resource
- 23:29where you can use the sheet in order to
- 23:31access sheet data as the resource. And
- 23:33it has prompt templates including
- 23:34analyze sheet data which is template for
- 23:36analyzing sheet data. Creating report
- 23:38template. So a template for creating
- 23:40report spreadsheets and form to sheet
- 23:42which is a template for form to sheet
- 23:44workflow in the code itself. Not going
- 23:47into too much detail over here. You can
- 23:50see that after we define the credentials
- 23:52the way that you can define these tools
- 23:54that you can decorate the function. So
- 23:56you know that's MCP tool. So this one is
- 23:58the function for listing the
- 23:59spreadsheet. This one is for reading the
- 24:01sheet etc. And then the way that you can
- 24:04show that it's a resource is that you
- 24:06just decorate it so that you know that
- 24:07it's a resource. So get sheet resource.
- 24:10And here are the prompt templates
- 24:12themselves. So they're actually
- 24:13hardcoded directly in these different
- 24:15prompt templates. So let's go back to
- 24:17claude here and see what is happening
- 24:19there. Great. We see that it analyzed
- 24:22the anonymous feedback spreadsheet. So I
- 24:25read that information, analyzed the data
- 24:27using the template that we provided and
- 24:29it actually created a dashboard as well.
- 24:31So here is the dashboard. So total
- 24:34response is 43. The average MPS score
- 24:37which is the net promoter score is 4.9
- 24:395.98 blah blah blah all these things. So
- 24:42it tells you information about this
- 24:45different categories uh how the study
- 24:47plan progress tracker is qualitative
- 24:50information about what attracted
- 24:52students to our boot camp and insights
- 24:55as well. So different types of
- 24:57information including what works well
- 24:59and what doesn't work well etc etc. So
- 25:01yeah there you go. This is the code demo
- 25:04of an MCP server. I hope you can see
- 25:06that there's so much potential for so
- 25:08many different types of things that you
- 25:09can build now that you have the ability
- 25:11to create your own MCP servers. You can
- 25:13of course also change your host. You use
- 25:14this MCP server with NA10 uh with your
- 25:18own agent with your any large language
- 25:20model application that you build as
- 25:22well. The possibilities are endless. So
- 25:24if you do want to dig deeper into this,
- 25:26I highly recommend that you actually
- 25:27check out the anthropic deep learning AI
- 25:29course that goes into the exact
- 25:31implementation step by step how to build
- 25:33this entire MCP server. It is really,
- 25:35really, really powerful. There are of
- 25:36course a lot of other ways that you can
- 25:37implement MCP servers as well, both code
- 25:40and no code approaches. I also really
- 25:41like this one by Network Chuck where he
- 25:44goes into a lot of detail about how to
- 25:45use Docker in order to do this. So I'm
- 25:47going to link his video below as well. I
- 25:48really recommend that you check it out
- 25:49if you're interested in implementing it
- 25:51using Docker. But for this video, I'm
- 25:53going to end it here. I hope that you
- 25:54now have an excellent fundamental
- 25:56understanding of MCP and how to actually
- 25:58build them yourself, too. Here is the
- 26:00final little assessment. Please answer
- 26:01these questions and put them into the
- 26:03comments.
- 26:06Hope you enjoyed this video and thank
- 26:08you so much for staying until the end.
- 26:09I'll see you guys in the next video for
- 26:11live stream.
- 26:17[Music]
About this transcript
This page contains the full transcript of MCP In 26 Minutes (Model Context Protocol) by Tina Huang, generated from the public captions YouTube serves with the video. The transcript has 5,477 words across 787 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.
What you can do with it
Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to translate the transcript into 100+ languages.
Free YouTube transcript tool
YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, translation to 100+ languages, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.