MCP Explained in 31 Minutes (Without The Hype)

now that open aai has officially offered their support for mCP it is basically confirmed model context protocol is going to be the aient framework today I'm going to explain what mCP is without the hype and without the fluff this video is suitable for complete beginners so even if you guys have never typed a line of code in your life today you're going to learn exactly what mCP is how it fits into your AI or automation stack how to use it in practice and I'm going to include a real example inside of naden my name is Nick I build systems for a living I actually scaled my own AI an automation company to 7 a month so this is literally all I do let's get into it so what is model context protocol I should note that it is in

every Creator's interest to make stuff like this whether it's a protocol or a framework or a software drop or whatever seem like it is bigger than it actually is okay so mCP model context protocol is just that it is a protocol and just like most protocols we can deconstruct it we can break it down we can learn it step by step that's what I'm going to be doing with you guys today the thing about mCP is that it's not actually a lot of people think it is cuz they're seeing it blow up all over the internet right now but mCP or model context

protocol was actually launched by anthropic way back in November 25th of 2024 so it's not exactly new it's also not exactly old but essentially the reason why you've been seeing it all over your YouTube right now is because it's finally gaining a critical mass of adoption where it's gotten to the point that it actually does provide a ton of value right like a protocol is only as valuable as the number of people that use it when it has wide Market penetration that's when you start to really see the benefits and because of a few creators or maybe anthropic Zone marketing efforts or whatever we're finally getting that point now so just to give you guys a little bit of context mCP itself is not new it's relying on technology that was created a year or so ago depending on when you're watching this video so to understand how mCP or model context protocol really works the best way is to just Dive Right In and build in my case I'm going to be using a lower no code tool called n8n just so we can see how the concepts play out in like a live practical scenario we're going to do all the Whiteboard or Blackboard stuff later so this is an

example of an NA an AI agent that is non mCP there is no model context protocol in here for all intents and purposes you guys can think of this as like the current way of building AI agents now I'm using NAD which is a low and no code platform if you guys don't have experience with this that's okay I'll walk you through it really briefly um essentially when we receive a new chat message in this window what happens is it calls our AI agent which is basically just a giant categorization machine it chooses which tool to call based on whether or not we ask for a tool and then if we don't then it just calls the open a chat model and returns answer and we also have a window buffer memory not super important to know the thing that is important to know is if we look over here we have all these tools right what I've done in this case I've added uh five I've added a tool for creating a calendar event one for deleting another for getting another for updating another for getting many and the key here and essentially the thing that mCP solves is mCP means that we will never have to go through and create all of these nodes and all their specifications ever again I'm never going to have to actually go into this n8na agent set the credential well I will we'll have to set a credential but set the tool description set a resource set an operation set a calendar set a start time set an end time and then set some additional fields in the summary I'm never going to have to manually configure each of these again okay and this may not seem like much but imagine you got like 10 or 15 different options for your a agent to categorize in orall um and you know if everyone takes 10 minutes that could be an extra couple of hours before you set up your AI agent flow right what mCP does is it eliminates this completely okay so this is like our pre-flow sort of like our current non mCP flow this is an example of an mCP flow okay and there are a couple things that look a little bit different here so um previously we had what like five different tools that were all Google Calendar tools right well now instead what we have is we have an mCP node one that's called execute tool another that's called list tool I imagine this is going to be deprecated soon we're just going to have one so I'll just keep it from a bird's eye perspective essentially we have um an appify mCP appify being a scraping platform and then we have an Airbnb mCP so notice how over here okay what we did is we had five nodes on one software platform well over here what we're doing is basically inside of apify is a giant list of nodes okay but we're only calling this one node and then it calls the specific node for us so this is sort of you kind of almost think of it like an a agent calling a bunch of other a agents which is a solution that people used to use in order to try and solve this problem but it's a lot more sophisticated than that and it just eliminates a large chunk of the issue um so basically what happens is like I'm going to open this chat okay and then now this is going to look a little bit different just cuz I'm on a different version and I got the um mCP community no installed and all that stuff so what I'm going to do is I'm going to ask it to tell me hey what

tools can I use with Airbnb and what it'll do is it'll call this list tools node which I set up in like 0. 5 seconds and then it's going to return me a list of all of the various tools that I can call okay so specifically Airbnb because this is a new mCP server only has two sub tools but I can do an Airbnb search then I have a list of all the parameters that I could feed in which will automatically be filled and then I have Airbnb listing details which are where you know I can get a specific piece of information for specific listing detail or something okay um how about the appify okay tell me about appify tools what it's going to do now is it's going to call the appify mCP server that appify mCP server is going to get a list of all of the various endpoints that it has access to and then it's going to return them and so you know just because again this is really new and early days stuff we only have one um specific tool that we can call called search but interestingly notice how before you know I would had to set up all this stuff um manually well this is now going to automatically input the query the Max results the scraping tool we're going to be using the output formats the request timeout and so on and so forth so basically I guess what I'm trying to say is in a nutshell instead of you having to connect to one tool like a create calendar event what you're going to basically get to do is you're going to get to combine all of these into a family of tools and then you're only going to have to connect that one family do that one authorization and then that's it you're good to go and as I've mentioned before there's an execute and a list here um I imagine that's just going to be combined into one in the future essentially what this is going to mean is you will pretty soon have an AI agent that is capable of calling a million things okay and more importantly if I go over to the giant list of mCP servers which is provided by anthropic and has a bunch of community servers so this is where you actually get all servers from what's more important is there's actually a service and they're developing these right now called mCP Compass mCP create whatever where you will pretty soon be

able to instead of connect a specific node that connects to a family of tools and then calls the specific tool properly you're actually going to be able to connect to a node called like mCP Compass or something some sort of high level node which will actually select the family of nodes that you need and then select the specific sub node of that family of nodes so we're basically just increasing levels of abstraction C here and if you think about it logically that's all programming and no code and AI have ever really done so to make a long story short right now if you guys want to build an AI agent in naden or I don't know zapier or one of these other lower no code platforms and you want to do something useful like grab data from a database or maybe send an email through Gmail or create a task and clickup whatever the hell your use case is you kind of have a problem every one

of the connections that you create will require a separate node or a module now every time you create you are obviously going to have to configure it right in NN that means that you guys are going to need to do like a poster SQL node you'll need to do a Gmail clickup node and basically all of these things are going to have their own unique parameters they're authentication methods and they're also going to have their own unique data mapping requirements the point I'm making is it's a ton of manual configuration um just to make sure that your workflow flows and if any Services were to ever update their apis the entire workflow would basically break until you go in and fix that configuration right I'm sure we've all been in that situation before is people that work with AI Automation and code and that's really the value prop of mCP it's basically a standardized way for AI agents to connect to external services without you having to manually configure every single connection instead of you having to set up a specific node for every integration what you do is you just give your AI agent the entire family of nodes the mCP server and it will just automatically at runtime initialize and have access to all of them so that's the bird's eye view let's now dig a little bit deeper into the weeds and cover some of the more technical stuff that he might have been confused watching other videos at okay so here's more or less how AI agents

work right now okay as you saw with the NN example we start with some sort of chat message and that goes to our AI agent right over here that is a really crappy line let me just rearrange this a bit after it's made it to our AI agent um the AI agent sort of does some thinking okay this is just like its own reasoning process with our wonderful little robot and then you know It ultimately chooses um essentially which tool to use based on that so I don't know let's say my new message is hey can you add XYZ to the database right pretty simple message what is this a agent going to do it's going to see the word database it's going to look at the tool that has access to which is database and it's probably going to go this way I say probably because as I've mentioned a couple times before these agents you know while they're smart and they're intelligent and they're great um they're not that smart and that intelligent and that great right now like on average uh you have an error rting anywhere between maybe 1 to 5% per request so kind of makes for like a rough go if you're a business owner and you know you just want a chat bot that you could just message and say add XYZ to database well I mean if it just doesn't do it 5% of the time you're losing more than 5% of money on it which is an interesting point somebody in my channel made a while ago anyway I guess the point that I'm making is you have access to all of these independent tools right so we need to set up the call database tool on its own postgressql or whatever we need to set up like the server spec we need to like set up all the the inputs we need to like drag and drop and map the variable whatever we saw that same thing with the send an email right create a clickup task right and Google Calendar appointment these are all different tools and so the AI can call them and it can do like a pretty good job but you as an AI automator and AI integrator um you have to like go through all the rigma r of dragging and dropping and setting them up okay how mCP AI agents work is a little bit different notice how we've now added a couple of additional layers to this so we still have a new message that comes in okay so I don't know you know add XYZ to DB okay but what we'll do is essentially when at initialization at runtime what'll happen is you see how this mCP client is connected to a bunch of these servers well basically what happens is and I can't really show this without just screwing up the graph so bear with me but at initialization these uh Services okay will feed back all of the functions that they have access ACC to which you'll see in a second to the mCP client will inject into the prompt of the AI agent a giant list so it'll be like hello you have access to post gresql you'll have access to email you'll also have access to CRM for postr you'll have access to function one function two function three for email right and it'll do so in this super compressed and pretty efficient format and now this a agent will see Okay add XYZ to database it'll be like okay well it's probably postgres because postgres is a database then it'll be like hm okay how do you know what's the add function it'll find that one and then it'll tell the mCP client hey I want post cres ql and I want like the add function okay and here is my input parameters MCB client will take that information and now because this is a level of abstraction it's like a level of distance basically between the llm and then the tool that it's calling this mCP client can format it in like a safe and secure way that is free of I don't know like prompt injection issues or like direct server access issues or or you know safety issues with the tools okay and now this mCP client which is you know it's not like an AI it's just a procedural logic takes this might be like the NAD chat boox window and then it'll actually go and it'll call the specific thing it wants so you know it sees post SQL okay great goes to the post SQL server and then what does it see if I scroll down a little bit more notice how this postgressql server mCP server now has access to and can call all of these endpoints and these are basically like kind of think of them as like API calls right so what's it going to do it's going to add a record so um I just realized that I don't think I included an add a record example that's really funny so we are just going to go down here and go add a record which was totally there before and then um it's going to you know it's going to choose the add a record then from this we're going to get the data right the server will get it it's going to go back to the server from this it'll go back to our mCP client then our mCP client will basically be able to again you know inject it um and kind of append it to the end of the agent thinking chain which will then send it back over to us and it'll say you know congrats our thing was successfully done okay great so now that we've seen this from a bird's eye view why is this you know why is this actually cool like important well notice how up here we had to do the input spec of every tool call DB send email crate click up task create Cal Point whatever like all of these were fundamentally different things that we were asking for and so we had to go and we had to drag and drop the variables and map them all independently and stuff what happens over here is that's handled by a level of abstraction you no longer need to do that all you need to do is you just need to authorize you need to connect to the poster SQL Server okay you just need to do your API key or or whatever the specific connection mechanism is depending on the software the second that you're done with that now you're good you don't need to do any additional input schema mapping of variables you just said whatever you want to the AI agent okay this AI agent will call everything else and it'll do all of the work for you and AI agent is probably not the right term that I'm thinking about because a agent is really this entire thing what I mean to say really is your large language model okay so your llm just does it all it it has all of the access to all of the input specs it just does it all automatically it's able to choose intelligently and here's the really cool thing it's more accurate it's substantially more accurate it's more accurate because all of the stuff is now standardized all of this stuff has been being worked on by people all over the world the prompts that we're using aren't just like a prompt that one engineer at NN or one engineer at some other company whipped up kind of you know in between their smoke break and just think like it works kind of well no these are actually like constantly being tested constantly being iterated and constantly being improved and then this sort of approach even though you know they're more steps and it's more refined and stuff like that takes that accuracy metric that I keep on bitching about from you know something like that maybe it's like 95% or so to much much higher it does so while um ensuring some level of security then ultimately speaking this is just like kind of the way that all things go right things as they improve the new products they tend to get a little bit more complicated a little bit more sophisticated but at the end of the day a little bit more accurate and reliable for users so model context protocol is composed of three main components we have the mCP client

which sits with your AI agent in NN or whatever platform you're using and you can think of this as like the software that wraps around the chatbot when you make a request via your large language model what happens is actually not sent to the service directly it's sent to the mCP Client First which will process it and then sanitize it before sending it to the next step which is the mCP server now the mCP server is what extracts or receives the request from the mCP client and these requests are formatted as a specific tool call with parameters like number of rows or search URL or whatever basically it takes this tool call and then it turns it into an actual API request that it then sends to your external service finally you have the actual external service itself this is the database or the Emil or the CRM or whatever and so what happens at runtime is the mCP server which has a big list of all of the available tools to call will actually compress and extract that down into a very tight efficient list that it'll then send to the mCP client then the mCP client at runtime will inject this big list into the prompt of the large language model itself when you tell your AI agent to update that customer record and clickup for example you now have your request but then on top of that request you'll also have that giant list of all the tools that the AI can call and that'll be written in a very standardized and a very efficient format your AI will then read the list of tools then read your request and then if it needs a tool to fulfill a request it'll make that decision extract what you want package that tool call into the format that the tool wants then it'll send a little snippet of text to the mCP client will take that request and send it to the mCP server then ultimately the mCP server will call your thing uh I know that is a lot of steps but there are a couple of upsides to this the main upside is by insulating an abstracting everything out like we're doing if somebody were to hypothetically build an mCP server for a service like clickup or air table or whatever what that would mean is anybody could use that with any AI agent that supports mCP which means mCP in that way will grow like a weed if you wanted to build a solution and you wanted to open source it or whatever you wouldn't just build it for yourself you would literally build it for every other human being and AI agent on Earth you building this mCP server or this solution or whatever would literally be adding to the collective knowledge of humanity and I think that's dope as hell another point I think a lot of people Miss is that mCP standardizes the language it standardizes the input schema it improves security because you're abstracting away the large language model and then the tool call right we have the mCP client in between right now basically every service has their own unique prompt system their own unique approach to security and their own unique you know format for input schemas and whatnot and a lot of these suck like if you've watched my channel before I constantly dunk on NN I rag on andn which is funny um I think that their prompt engineering for tools sucks so the whole idea is screw that we as the entire internet will move towards a unified standard and then as a result everything will just get way better it means that you won't need to manually configure all those connections maintain them when apis change um mCP basically virtually guarantees that everything will just be plug andplay instead of the node configuration hell that I think we've all experienced ourselves so now that you understand what mCP is let's talk about why that might actually matter for people that build aom people that make money because at the end of the day this is anational Channel and I want you guys to know everything that you need to know in order to do work but I don't just want to focus on the knowledge if we just get down a brass taxer the whole point is money right so

how is this going to help you make more money well think about the current situation if you guys are using naden or make. com right now or whatever other AI agent tool you're basically having to build out every single node as well as the input specifications of that node you'll have to drag and drop various things into various fields and have to go dollar sign from Ai and that works fine for simple agents but it gets complic as hell when you try to build anything bigger CU you're basically going to have to recreate the wheel every time and when you do that you'll have to recreate the wheel for like a massive agent with a million tools the ones that look like spaghetti diagrams which I think we've all seen the thumbnails for mCP solves this instead of having to hardcode every possible tool basically what happens is your AI agent will just connect to the entire Suite where the entire family is upet which is just the mCP server and then it'll get a bunch of context about which tool it needs to use so this is something that people uh do in the space pretty often right I always talk about League process in lead enrichment lead extraction that sort of stuff so hypothetically as an example if a new lead were to come into your system right now you might have to build dedicated steps to first check it against your database after you you might do a little bit of research on the company scrape the website find the email address you know enr a field or whatever right that's what you have to do right now you have to build out this logical left to right flow and make. com and it's mess pain in the ass with mCP what you could do is you could just add an AI agent node you

could connect it to an mCP server called like your lead enrichment server or something and basically what'll happen is when a new lead comes in you just say hey could you research them add context around who they are and what they're doing to the CRM and then assign them to the appropriate rep you as an AI automator that's all you have to do you no longer need to go through this whole rig Ro of like putting together this procedural logic it'll just do it right and you don't even really need to go through the whole riger roll of like connecting any of this stuff because you just connect to one server which maybe it's connected to a bunch of other servers or something I don't know but it'll just do it all for you so you know you connect it to that one tool like maybe a search engine or some lead instantly thing you know that does it all for you then it'll go through the whole thing for you it'll scrape the lead it'll enrich it for you it'll do everything you need and because mCP is standardized as hell it'll do so with substantially higher accuracy and reliability which means it'll actually start being able to use it for business applications so yeah you know here's the key thing that people are missing it's not magic it's just a protocol it's a slightly better alternative to what most people are doing right now it's a way for different systems to talk to each other if you guys use nadn for instance all of what I have just Des described is more or less what n agents are doing right now it's just they're it's kind of harder and more annoying mCP is just like one level of abstraction up it's just a more efficient and global standard for how to lay things out when you ask your a agent something it's not like it magically knows what's going on at the other side of the flow um as I mentioned at runtime the mCP server model context protocol this protocol is just hey list all the endpoints that you have access to inject that into the large language model via the mCP client and then voila right it just does it a little bit better than most approaches and that's really the cool part about protocols when enough people get on them and we get to the point where we can improve them a little bit by little bit everybody's experience improves at approximately the same rate which is sweet so what I mean by this is we're not fully there yet not everybody's using this not all the tools are there so you know the implementation matters the quality of your llm still matters the specifics of how each mCP server still matters a lot um but you know I want you to know that like the goal is to ultimately move forward to this um a standardized way for systems to communicate and uh you know one where we don't necessarily need developers to do every single tiny little connection we just need like hyper effective developers to set up these MCB servers and then everybody else around the world gets to benefit so let's look at where mCP is at right now I spent some time

testing various mCP servers with different AI agents and just to be brutally honest with you we are still very early right now there are basically three main ways to use mCP and these mCP servers are kind of Hit or Miss um but I will explain them all to you regardless

first you could use mCP with a claw desktop app anthropic is the company that created the mCP standard so naturally their implementation is the most secure you can connect local mCP servers to CLA and have it interact with your file system or your database or some basic web services or whatever uh the second way is with development environments like cursor AI I personally think this is where they have a lot of value right now um I've seen a lot of practical use cases just on Twitter and stuff like that cursor can currently use mCP to understand your codebase search documentation and even generate code that works with your specific project structure as well as like call some tools that you know you might want and you might have previously had to add some hardcoded Logic for um and then third you can actually build your own implementation using the open source mCP standard in library now I wouldn't personally do this as an AI automation person because I think that if you go through the whole rigoll of like building out all that stuff it kind of defeats the whole purpose of using these no code and automation tools right which to me is to avoid heavy development work but then drive disproportionate outcomes in terms of Revenue um so like the reality is a lot of people that do this mCP server stuff on their own time maybe they're not actually that good right now they're still pretty experimental pretty buggy or pretty Limited in what they can do um but you know eventually we get to the point where these are like production ready and you can use them in your day-to-day um but yeah you know that's kind of like the current state there's no robust Salesforce mCP server yet there's no Google workspace mCP server yet uh there is no real integration for most popular marketing tools right now uh but there probably will be right for mCP to realize this potential and I think now with open a behind it will we do need those high quality server implementations for every service we want to connect to and that's just going to take time just think about how long it took for every SAS product to have a decent freaking API and like so many of them still don't right now Imagine The Complex creating mCP servers for all of those apis plus all the ones that don't have apis and it's not just about setting up like an endpoint now it's setting them up in a way that an a agent can connect to an a agents are smart but obviously there are certain limitations surrounding them so I guess the point I'm making is tldr there's a lot of work so where does all that leave us as automation Builders and agency owners are there actual opportunities here is this just another AI hype cycle Nick what should I do I see two

realistic scenarios for how you might want to approach mCP right now the first option is just wait and see if you guys are focused on delivering Client Solutions that work reliably today that's probably the most sensible approach mCP is not the shiny object okay it's not the shiniest object or whatever and it will be valuable but right now it's kind of a little shiny it's still maturing and betting a real client project on mCP is probably not the best for the client interest the businesses that you're working with generally more proven technology um I don't think we should be you know helping them install buggy experiments into their flow you know I think it be great eventually it's just going to take some time and yeah that's kind of like your first option that's you know I'm doing a little bit of that most people in my community are doing a little bit of that and so on the second option is for you to start experimenting internally um I'm also doing a little bit of this you know I'm testing different mCP servers inside of nadn specifically I'm seeing what works well and what doesn't and I'm thinking about how these could eventually fit into workflows that I build um and basically I'm not like implementing these in client projects I'm just building expertise that might be valuable later so you know the real business opportunity I think that we can get out of this as Ai and automation people is not necessary to try and tie an mCP directly into all our client projects right now it's just to position ourselves as people that understand the direction the technology is headed um we can use this as Authority we could use this as like a sales tool uh you know we could use this as like future potential also speaking of potential I think there's like a big opportunity right now to build highquality mCP servers for popular business tools probably over the course of the next month so if you guys got some development skills or maybe you're a company that has access to developers like you know I was just in the shower thinking about this so screw me if it's not exactly the most forward or well thought out business plan but you could probably create mCP servers for tools that don't have them you could probably approach these businesses and offer to do that um so that you know these businesses get to take advantage of all that AI agent traffic I think there's going to be a ton of demand for it um and as I mentioned with open AI now behind it the two biggest AI companies in the World Behind mCP is virtually guaranteed that this is going to turn into something much bigger than it is right now all right so let's get practical for second how does mCP actually fit in with the tools you're already using like naden if you guys are using naden you might have notice that they recently added mCP capability through what's called a community node this is interesting because it starts to bridge the gap between these structured workflows I was talking about and these more flexible a agents that we were just

discussing um it's still pretty early days to use the mCP node in nadn you do need to be running a self-hosted version not a cloud version you also need to go through some rig roll like you have to enable Community nodes you have to set up your environment variables correctly and then you have to install the mCP node manually which means it's not plug-and-play um the reality today is if you try to implement this for a client or for yourself you probably spend more time wrestling with technical issues and actually delivering value like it took me a good couple of hours to actually figure out what was going on with this but you know that's the vision and I do see the market and N end moving towards that for sure I was just browsing a forum thread like on the in and Community forum and I saw that the developer said like oh hey you know we're removing the requirement for you to have like our community underscore nodecore tool underscore usage environment variables because we think people are going to be using mCP like crazy so I mean really it's just a matter of weeks and or months before naden has access to this sort of stuff and who knows by the time you're watching this maybe anden has like mCP as part of their core functionality I don't know so yeah for most of you guys running client Focus automation business

definitely follow this as interesting just don't build it for a client yet it's worth understanding conceptually maybe even experimenting with a little bit internally but probably not something I would use for a client project just yet an interesting side note I'm going to make is that mCP is basically just an AI wrapper around apis you take this API specification and that

API spec was always built for humans initially right API end points and docs and all that stuff these are human things but then what AI allows us to do is we basically aify the whole ecosystem of apis and turn it into mCP servers so if and when mCP continues to grow in the future what I think is going to happen is like every API will be developed alongside an mCP server because that's actually going to be much more important to the internet than just us lowly little humans like you know if there are eight billion humans n billion humans however many billion humans there are right now that we know about um you know I think there'll be like 90 billion AI agents and so obviously like the internet will have to be made for these AI agents right more so than the human beings and our py little uh API end points where do I see all this stuff going over the course of the next 6 to 12 months first I think we're going to see a flood of mCP servers being built for the most major past providers to start and then that'll just trickle down to all of the um you know analoges and stuff like that so think of all the platforms I use on my channel click up Panda do big crms like Salesforce or HubSpot or QuickBooks Google workspace the dreaded Microsoft 360 Suite I think all of these will eventually have dedicated mCP servers that let AI agents interact with them naturally I don't think they're all going to be good I think a lot of them are going to be buggy and kind of crappy and I think we're going to have the same problem we've always had with new tech implementation I think the quality will vary widely but it will get better with time I think that'll be in like the next three to six months and then I think you know as time goes on we'll see mCP become a standard feature in these AI development platforms so curser um you know Claude Devon whatever you know I think all of these AI development platforms are going to support mCP as basically foundational and then um you know since open AI has now expressed support behind it I think this is now going to trickle down from infrastructure to actual tools so what do I mean by tools I mean naden make. com zapier all these other no Code and low code Builders so um we're already seeing this with naden with their experimental support that is slowly transitioning into core features but um yeah you know I think that it's just a matter of time before make. com picks it up lovable um Lindy basically any one of these platforms um just starts using that as like a core feature so does this mean anything for your business really probably not for the next few months still too immature to build production systems on but I do think in the medium term uh we'll probably start seeing some opportunities to use mCP how fun is that um really excited so to wrap it all up I do think

mCP is a great step towards making AI agents genuinely useful for AI automation I think it addresses a real problem which is the difficulty of connecting AI systems where our actual data business lives to you know real applications I think that another big issue is accuracy and I think mcps are going quite away to improve the accuracy and reduce like some of the weird one-off errors that we see with particular implementations like n but to hedge at all we are still in the early days there's a pretty big gap between the hype and the current reality I think most mCP services today are kind of unstable they're kind of limited in functionality and you know I don't think this is going to be a hype bubble that just disappears I don't think this is just like a short-term thing I do think it's worth paying attention to but I think it's going to take a little bit more time and I think that uh you know this critical threshold of interest in MC and stuff like that you know it's probably a little bit still too early kind of the same way that N na agents were just a little bit too early and they didn't really like actually in actionably help uh business outcomes I think that's kind of like where we're at with that and that's fine just wait a couple months jump in a hyperbaric Time Chamber get in a time machine or something when you come out mcps will be everywhere and that's that if you guys want to stay updated on a automation including future developments with mCP and maybe some other tools just subscribe turn on notifications I'm constantly testing these Technologies and separating the hype from what I think actually works for those of you guys that are serious about making money with AI I would encourage you to join my community maker school we have over 1300 committed AI entrepreneurs that are growing their AI agencies you can get that in the description below aside from that leave a comment down below if you guys have any thoughts or wants or questions I really appreciate you guys watching and I'll catch you on the next one take care