🔗 Subscribe to our Newsletter!
Get the latest AI updates, tips, and insights straight to your inbox:
https://dub.link/staying-ahead
I ordered real pizza by just talking to my computer. No apps. No clicking. Just one conversation.
This is MCP (Model Context Protocol) - the technology that lets AI interact directly with your apps. Zomato, Spotify, Gmail, and hundreds of other services now work this way.
In this video, I show you:
- Live demo ordering food through AI
- How to set this up yourself (Claude & ChatGPT)
- Why Microsoft, OpenAI, and Google are betting billions on this
- The security risks you need to know about
This isn't the future. This is happening right now.
And if you're a marketer, creator, or business owner - you need to understand this before everyone else catches up.
--------
To Know More,
Follow Vaibhav Sisinty On ⤵︎
Instagram @VaibhavSisinty
https://www.instagram.com/vaibhavsisinty
Twitter @VaibhavSisinty
https://twitter.com/VaibhavSisinty
Facebook @VaibhavSisinty
https://www.facebook.com/vaibhavsisinty/
LinkedIn - Vaibhav Sisinty
https://www.linkedin.com/in/vaibhavsisinty
Look at my screen. I'm not going to click. swipe. I'm not going to open a single app. But you are about to see me order a real pizza with real money. All from this one chat window. Watch this. Order my usual margarita pizza from Zomato. Done. Real pizza, real money. Food is literally on its way. It's incredible, right? But also, think about what just happened. This AI just confirmed my home address, processed my credit card, and knows my favorite food. It has my data. That's a little creepy, right? But that's just the start. This new MCP protocol isn't just for one app. It's a key that unlocks all of them. Watch this. Spotify, I need a playlist for deep work. I like piano and ambient. Nothing with lyrics. Should feel contemplative, but not sad. About 90 minutes. A brand new custom playlist made in seconds. It didn't just order my food. Now it knows my music taste. So this AI has my location, my credit card, and my personal taste all in one place. But here's the real loophole, the one that should make you nervous. If this AI can use all that data, what's stopping someone else from asking for it? I spent 48 hours testing this, and what I found is a vulnerability that is terrifyingly easy to exploit. Let me show you how this magic is also a massive security flaw. So, I was scrolling through GitHub last week and I found something interesting. Zomato, yes, the food delivery app had released something called a server that lets AI talk to their platform. I'm searching for it now. Here it is. The page explains what it can do. Browse restaurants, check menus, place orders, all the stuff you'd normally do by tapping through their app. And there's this section here for Claude desktop. That's the AI assistant I use. Let me copy this configuration. Switching to my claude app. Now I go to settings, then developer, then edit config. This opens a file. I'm just pasting what I copied. Save. Close. Now I restart Claude. This is the crucial part. When Claude starts back up, it needs to see this new connection. Let me check if it worked. Going back to settings. There it is. Zumato is now listed. It's running. Claude can now talk to Zumato systems directly. That's it. That's the entire setup. Maybe 3 minutes of copying and pasting. All right, let's see what happens. I'm going to type exactly what I'd say to a friend. Suggest me some nice veg Italian restaurant. Look at this. It pulled my address from my Zumato account. It's asking where I want delivery. I'll click my address. And now it's showing me restaurants, not just names, ratings, delivery times, distance, everything. Notice something. I didn't describe my location. I didn't specify a neighborhood. Claude already knows because it can read my Zomato profile. Let me try something more specific. I'm asking it to show me the menu from one of these places. Look, it's pulling the full menu, prices, descriptions, photos. Now, watch this. I'm just typing order this. It's building my cart, showing me the total, asking for confirmation. I confirm and done. Order placed. It's right here in my actual Zomato app. Order confirmed. The restaurant accepted. Food is being prepared. I just ordered lunch by having a conversation with my computer. And as you can see, here is my order. And that's my friend receiving it for me. So, what actually happened here? Think about how you normally order food. You open an app, you scroll through restaurants, you tap into menus, you read descriptions, you add items, you search for offers, you go to checkout, you select payment, maybe eight or 10 steps minimum. What used to take multiple taps, scrolling through menus, and navigating checkout just took a single conversation. But more importantly, Claude remembered things about me, my favorite area, the types of food I like, my usual order size. The app treats every order as a fresh start. Claude part of an ongoing relationship. This is what I mean when I say something fundamental has changed. Here's where it gets interesting. Zumato isn't unique. I found dozens of these servers on GitHub. Zeroda built one for stock portfolios, Razer Pay for payments, GitHub for code repositories, Slack for team messages, Google Drive for documents. Each one lets AI interact with that service directly. And then in March, something huge happened. OpenAI CEO Sam Alman posted on X. People love MCP and we are excited to add support across our products. Wait, what's MCP? MCP stands for model context protocol. It's an open standard introduced by Anthropic, the same company who owns Claude in November 2024 that standardizes how AI systems connect to external data sources and tools. That's the technical name for what I just showed you. But forget the jargon for a second. Here's what matters. In March 2025, OpenAI officially adopted this across Chat GPT. In April 2025, Google DeepMind confirmed support in Gemini. Every major AI is now learning to do this. Now here's where it gets interesting. Last month I was at
open demo day in Bangalore and the Zomato team came on stage to demonstrate something incredible. They open chat GPT and ordered food live on stage just by typing. The audience went completely silent because what they showed required zero technical setup. In October 2025, OpenAI launched something called apps. It allows ChatGpt to interact with apps like Spotify, Booking. com, and yes, Zomato. Here's the difference. With Claude, I had to go to GitHub, find servers, edit configuration files. It's powerful but technical. With Chat GPT apps, click add, authenticate, done. No GitHub, no files. Anyone can do it. Here's what clicked for me. Both approaches use the same foundation, that bridge system I showed you earlier. Think of it like Google Translate, but instead of translating languages, it translates between AI and apps. But OpenAI made it simple enough for anyone to use. This isn't just for developers anymore. This is for everyone. So, naturally, I had to try it myself. We've all been there. You want a very specific playlist. You open the app, you type focus, you get a bunch of results, you type piano scroll, ambient scroll, then you have to manually find one with no lyrics. It's a hassle. But now we can just tell it what we want. So I'm right inside the chat GPT app. I'm not even opening Spotify. Watch. I'm typing a really specific natural language prompt. Spotify. I need a playlist for deep work. I like piano and ambient. Nothing with lyrics. Should feel contemplative but not sad. About 90 minutes. Look at that. I'm giving it a vibe, rules, and a length. You can't do that in a normal search bar. It's talking to Spotify right now. And boom. It's generated a brand new custom playlist. I'll just tap the link. And here we are. Let's check it. It's all piano and ambient. No vocals. And the length, 1 hour and 32 minutes. It nailed every single part of my request in seconds. No more endless scrolling, just conversation. And that's the breakthrough OpenAI made. They took the power of MCP and made it accessible to anyone. And they're not alone. In April 2025, Google DeepMind confirmed support for MCP in Gemini. They didn't just adopt it, they went aggressive. Google announced partnerships with Workspace tools, Gmail, Calendar, Drive, Docs, Sheets. Imagine this. You're in a Gemini chat and you type, "Review next quarter's plan for the launch of our new shoe line, Strideex, and research current industry trends to propose any changes or improvements to our campaign strategy. " Gemini includes sources from your workspace and other external sources to complete this research all in one go. Microsoft has copilot integrated into office. Google is integrating Gemini into workspace and both are using MCP as the foundation. The three biggest AI companies, Anthropic, OpenAI and Google are all betting on the same protocol. This isn't an experiment anymore. This is the foundation being laid. But let me hit pause for a second because this isn't all sunshine and productivity gains. When OpenAI launched chat GP apps, they labeled it powerful but dangerous. Think about it. If AI can order your lunch, it can also send your emails. If it can send your emails, what happens if someone tricks it into sending the wrong email or deleting important files or transferring money? In May 2025, researchers found a critical vulnerability in GitHub's popular MCP server. The exploit could trick AI into leaking private repository data. Imagine that with your ad accounts, your client database, your financial data. In July 2025, back/security found widespread security issues in public servers, including something called neighborjack, which dangerously exposes systems on local networks. They're warning. When network exposure meets excessive permissions, you get the perfect storm. This power comes with responsibility before you connect everything to AI. Understand what permissions you are granting. Only connect to servers from known and trusted sources. Review what data the AI can access. Test with non-critical accounts first. Keep backups of important data. The companies building these connections are moving fast. Security is catching up, but you need to be smart about what you connect and when. Don't give AI the keys to your entire digital life until you understand exactly what doors those keys open. So, with all these risks, you'd think companies would slow down, right? Think again. At Microsoft Build 2025, their CTO Kevin Scott said something that stuck with me. — So, MCP is kind of like HTTP. Um, it is a protocol where you can connect to a service or you know someone who's got a piece of content they want to make agent accessible. HTTP made the internet possible. MCP is making the AI internet possible. Look at who's betting on this. May 2025
Microsoft announces Windows 11 native support for MCP. June 2025, Square, Stripe, PayPal, and Web Flow all demonstrate MCP integrations at Cloudflare's developer conference. October 2025, OpenAI launches chat GPT apps built on MCP. Today, Block's entire engineering team works this way. Stripes developers are building on it. PayPal is shipping it. The question isn't whether this will happen, it's how fast. And the marketers, the creators, the professionals who figure this out first, they're going to have an unfair advantage because while everyone else is still clicking through apps, they will be having conversations with their entire tech stack. So here's my challenge to you. Set up one connection this week, just one. If you have chat GPT plus, add an app. Google calendar, Gmail, Spotify, something simple. If you're using Claude, find one MCP server on GitHub and set it up. Get comfortable with this now. while it's still early. That's it. Subscribe and I'll see you in the next one.