You run a SaaS or web app and want zero-friction distribution directly inside Claude or ChatGPT.
You want a practical walkthrough of setting up and hosting an MCP server on Railway using Claude Code.
You are preparing an MCP connector for Anthropic's official directory and need to pass authentication review.
SKIP IF…
You are looking for prompt engineering tips rather than backend protocol implementation.
You do not have access to developer tooling like terminals, cloud hosting, or API endpoints.
Free for members
Chat with this breakdown — free.
Sign in and you get 23 free chat messages on us — ask for the hook, quote a framework, find the exact transcript moment, generate a markdown action plan. Bring your own key when you want unlimited.
“It's basically the USB C connection portal. for your AI agents that can talk to any tool in a very easy way for them.”
Clear physical analogy describing the abstract concept of Model Context Protocol.→ IG reel cold open↗ Tweet quote
00:03
“turn your app into an agentic first application so that apps like Cloud Code and Cursor and Codex can actually use this tool, which basically puts your product in front of all of the users of these platforms.”
Concise summary of the strategic business value of building MCP connectors.→ newsletter pull-quote↗ Tweet quote
The Script
Word for word.
Read-along
Don't just watch it. Burn it in.
See every word as it's spoken — crank it to 2× and still catch all of it. The same dual-channel trick behind Amazon's Kindle + Audible.
17px
metaphoranalogy
One of the biggest marketing hacks in the AI era is to turn your app into an agentic first application so that apps like Cloud Code and Cursor and Codex can actually use this tool, which basically puts your product in front of all of the users of these platforms. And for example, like my agentic social media scheduler, Creator OS just built an MCP, and that's what allowed us to connect our Creator OS app to Cloud Code and to Cloud.
so that anyone inside of cloud can just connect it directly inside of their chat, like you're seeing on the screen right here, it'll then redirect to your application, you hit allow, and all of a sudden, your customers are using your product inside of cloud. But for that to be possible, you have to build an MCP server and MCP server is a model context protocol.
It's basically the USB C connection portal. for your AI agents that can talk to any tool in a very easy way for them. And so the beauty of MCP for us as entrepreneurs is that we can build tools and skills and workflows and agents inside of this server.
And it can only be unlocked if they're using our apps. And so in this video, using cloud code and cursor in the next five minutes, I'm going to show you exactly how to build an MCP server from scratch. And today's video idea is basically going to be a shared notes brainstorm.
So imagine you were working in a team, and there's like 10 of you spread across the country, where you can build an MCP server to brainstorm ideas, and then you can have it in a knowledge graph type of way. So that you know, all of your ideas are being ranked against one another and creating connections between one another and so on and so forth.
But the reason why I'm showing you this example is so that you know how to set up your MCP server so that when you're building your real applications, you can also get it on cloud like I did for my app. So now I'm going to walk you guys step by step on how to set this up. The only tools that you need to get this working is some type of cloud provider where your MCP server is going to live.
In this use case, I'm using Railway because for me, it's what I'm using all of my apps to run on. And you need to have some sort of database to store the data if your MCP server needs that. So any Postgres database works there.
I'm using Insforge. And now I'm going to walk you guys through the prompt by prompt setup with Cloud. The first message here is basically saying I want to build a note taking app that could be shared between a bunch of members using an MCP server.
Essentially, we want to build a note taking MCP server. And then we tell it to one shot. After that prompt, it'll tell you what its game plan essentially is on how the infrastructure of the app is actually going to run through the MCP.
And from here, you want to give it the actual tools that I mentioned. So you want to give it the API keys to your database, the API keys to your cloud server so that it can go and get everything set up end to end. And so from there, we were able to get connected to Railway and we were able to see on the right hand side that it's called the Team Notes MCP.
And what that allows the agents to do is then register the MCP server with a real domain. And so here we have our domain as team dash notes, MCP production. And if you're using like your real production websites to connect to your MCP server, you would do the same thing wherever your real website is hosted.
So if it's like GoDaddy, for example, you would just add that CNAME and the agents will help you through that. But from here, What I told the agents to do as the last step is to build some type of like nice interactive team knowledge graph UI, just to see that everything's working.
And now we can tell that it's working as well as we have a full team members like graph here. And the agent gave us the full setup instructions. So now we want to ask a couple of questions before we can verify that everything's actually set up is one talking with my cloud chat right now, can I send in a note?
And then two, if I open up a brand new cloud chat or cursor environment, I'm able to connect to this MCP server. And how do I do that?
So first one, let's test if we can send in a message. So we're just going to say, you know, add in a simple message to knowledge graph right now. All right, so it wants to do it in a live way.
So apparently, we're going to see these. Apparently, we're going to see these messages pop into the knowledge graph. immediately.
Alright, so upon refresh, we have the Hello team message by the Kevin. So that was from that chat right there. So we can verify that that is working.
So the first step, the second step, let's ask the agent how we can get other people to connect to this server. So the agent let us know how to actually install it and get other people to install it.
So even you guys right now are going to be able to interact with this MCP server, I'll keep it up for the next few days. And basically, all I have to do is add in this command. And what's cool about this command is that at the end, we add in their name into the URL and their name is how the knowledge graph can then differentiate everyone from each other.
So now let's try that in a brand new cursor server. So we're going to open up cursor, we're going to hit new window caps, and I'm going to add in a brand new agent in a brand new folder that has no idea about this, I'm going to paste in that setup command, I'm going to say to be like, you know, say our name is Elon in the knowledge graph.
and add like five points to it. Here we go. So the agents now initializing team nodes, MCP, it's checking the identity.
So we are now connected as Elon. There we go. We can now see the contributors in the bottom right -hand corner.
So we have Jordan, Kevin, Leo, Maya, Priya, and Sam. So if I hit a refresh, we should now see Elon. And there we go.
Elon as the contributor. And now we're going to wait for the new knowledge graph to load because all of this new information has been loaded into the MCP server. And so we're going to give that a sec.
Bang, there we go. So now the knowledge graph has grown quite a bit. And we can see all of the green nodes are from Elon.
So we have an enterprise deal plan, a hybrid pricing proposal, all this nonsense. There's a couple of things that are stopping this MCP server that we just built from being added to cloud and being accepted. And the first one is that I was pretending to be Elon.
And so the next steps from here would be to add something called an OAuth sign in so that I would have to actually sign in with my real email address to gain access to this. And if I wanted to even make this private in general, that is another way where I can token gate the MCP server by email addresses, right? So these are things that you want to think about.
when you're playing around with this MCP server that you're building out. And so let me just quickly show you what that looks like for my MCP that's hosted inside of cloud, we go over to the Settings tab, we go down to connectors, we can go to the Discover area and search for connectors. Here, I'm going to type in creator OS.
And there we go. And so we can currently see that it is connected. But in this case, let's just quickly disconnect and quick.
And here, we'll click connect to cloud. And what we've integrated is this is this single sign on that redirects the user back to Creator OS so that they choose which workspace they actually want to enable, and then whether or not they want to have read or write access.
And then from there, the MCP server will officially be connected to the cloud. ecosystem. And in that way, there's no there's no way for anyone else to be signing in as me, for example.
So that, guys, is how you build the MCP server. And this is very important when you're building applications in this new agentic era, because everyone's using their apps through cloud and codecs and stuff like that. And after going through the full submission process and getting my MCP approved, I guess the points and tips that I can leave you guys with is that all of your endpoints should be tested by yourself first.
So you should build your MCP server, you should then connect it to cloud in your own environment, and then test every single endpoint to make sure it's all working so that when you do submit it for review, all of those checkpoints end up getting approved. And then before you know it, you're on the cloud ecosystem and Codex ecosystem.
My approval process literally took like an hour for this to work. So what that means is that their process is automated. And so if you make sure that all of your API endpoints are working, you'll then end up getting approved most likely, as well as having the OAuth sign in integration.
And from here, you can literally push your MCP server to cloud .ai and chatgbt .com. And you can literally have your apps running inside of their platforms where there's literally like a billion users.
The Hook
The bait, then the rug-pull.
The creator cuts straight to the core distribution leverage: turning existing applications into agent-accessible MCP tools exposes software directly to millions of active LLM chat users.
Frameworks
Named ideas worth stealing.
02:24model
The Agentic MCP Distribution Stack
Cloud Hosting (Railway)
Relational Database (PostgreSQL / Insforge)
Protocol Server (TypeScript Model Context Protocol)
The structural architecture needed to take an existing SaaS database and make it securely discoverable and executable inside AI chat clients.
Steal forplanning architectural migrations to expose legacy SaaS APIs to autonomous agents
CTA Breakdown
How they asked for the click.
VERBAL ASK
08:14link
“And from here, you can literally push your MCP server to cloud.ai and chatgbt.com. And you can literally have your apps running inside of their platforms where there's literally like a billion users.”
Pitches the strategic value of public deployment and points viewers to description links for Creator OS and developer communities.
Add Modern Creator as a preferred source and Google shows you more of our breakdowns in Search, Top Stories, and AI Overviews. It only changes what you see, and you can undo it in your Google settings anytime.
Add to Preferred SourcesOpens your Google source preferences with us pre-loaded. Tick the box and you're done.
Kev wires the tool-dispatch model Jev into his open-source editor HyperEdit, so a chat prompt triggers real cuts, captions, and media pulled straight from an Obsidian vault.
A walkthrough of Lauren Tan's pstack: 21 engineering principles, 22 playbooks, and 24 skills that turn a coding agent from a slop machine into a verification-obsessed engineer.
A walkthrough of a free GitHub tool that reroutes Claude Code through free-tier API keys from NVIDIA and other providers once the paid usage limit runs out.
A change-log walkthrough of five new AI-coding skills, built around one idea: let the agent handle judgment, but keep humans in charge of anything you can't undo.
00:09
00:15
00:21
00:27
00:33
00:40
00:46
00:58
01:04
01:11
01:17
01:29
01:35
01:41
01:48
02:00
02:12
02:18
02:25
02:37
02:43
02:49
02:55
03:02
03:08
03:14
03:26
03:33
03:39
03:45
03:57
04:03
04:10
04:16
04:22
04:28
04:34
04:40
04:53
05:05
05:11
05:18
05:24
05:30
05:36
05:42
05:50
05:55
06:01
06:07
06:13
06:19
06:28
06:32
06:38
06:44
06:50
06:56
07:02
07:10
07:15
07:20
07:27
07:33
07:40
07:46
07:52
07:58
08:04
08:10
05:17
12:57
11:13
23:07
11:37
15:09