MCP can have significant impact on habitual internet usage practices

MCP can have significant impact on habitual internet usage practices

Model Context Protocol (MCP) is now popular subject in discussions around AI and LLMs. It was designed to add a standard way to connect "external" tools to LLMs to make them more useful. Classic example is the "what is the weather in ..." too. Each AI chat tool could do this with own way. Now there is a standard and a plugin made for one Ai Chat system can work with others. We can se burst of enthusiasm in implementig of MCP servers for everything. I expect this trend will grow. Especially usage of MCP servers with SSE transport. Implementing of MCP server with Server-Sent Events make it similar to SaaS server designed for LLM/AI tool as a client. There are to reason i decided to write this artcile. - First. It is reported that internet users now often go to AI chat (often ChatGPT) to find something instead of going to google - Second. OpenAI anounced they will add support of MCP to ChatGPT Desktop soon. And they will add both STDIO and SSE transport protocols for MCP Based on this i expect we will see some interesting changes soon. ## User cases we can see soon ### Sell with ChatGPT (or any AI Chat) Imagine you have a small flowers business and you sell flowers with the website. A client goes to your web site, adds a bouquet of flowers to a card and then make an order. You can show your clients a button "Add this store to ChatGPT". When a user cliks your MCP endpoint is sent to ChatGPT "Add external tools" form and a client accepts it. Then a user opens the ChatGPT and writes: ``` User: I need to order flowers to my grandma ChatGPT: Today there are some nice offers in the store "Business Name". I think one of next three options should be fine. There are options: .... Or do you want to view more options? User: The ptions 2 is fine. Please oreder to my gnandma's address. ChatGTP: Some message on the card? User: Yes, decide yourself. ``` And the order is created. **What happens on background.** The flowers show has implemented MCP interface with SSE transport. THis means there is the endpoint on the web site , something like https://flowers_shop.com/mcp listening and accepting requests in MCP protocol. To implement this feature it is needed to do some minimal coding. because MCP server is just a wrapper over existent functionality of the web site. Once this endpoint is added to ChatGTP then on every communication with a user, ChatGTP will "keep in mind" all connected tools. And when a user mentions something that can be related to one of tools ChatGPT will understand that a tool should be used and will use it. The MCP server in this case will have following tools - browse, search , add to card, order etc. ### Expose data to ChatGPT (or any AI Chat) THere is the another example. Imagin you have a community board on the web. Good old forum. Every day new discussions are started. And our user likes to participate. BUt also, our user now spends more time using ChatGPT. This is his "first tab" in the browser. On your web site you can add new button "Add the forum to your ChatGPT". A user cliks a button and confirms on the ChatGPT web site. Now next time when the user opens chatgtp he can write ``` User: Whats new? ChatGPT: There are 5 new dfiscussions on the forum "Forum name" and one internal message. User: What is the message ChatGPT: The message from the user _user_. The text is .... User: Respond him i am not interested ``` **What happens on background.** When a user asks "Whats new?" the chatgpt checks all connected tools if there is any tool that is responsible for this. It find the forum MCP offers the function to get latest new. And it calls it because thinks it is the best option for now. Then ChatGPT interpretes the response and whos to the user. In there example MCP server will have to implement endpoints like "get_news", "get_internal_message", "post_internal_message" etc. These two examples show have any web site can be easily integrated with the ChatGPT or any other AI Chat system (i believe most of them will support MCP soon). And now imagine how this could work if Ai Chat programs will start to support "Speech recognition" to be able to talk with the client using audio. Patrially, this is already done. LLM + Voice communication + the set of favorite MCP servers will make the magic! ## MCP interface for any web site can be a "must have" soon A trend to use ChatGPT/Ai Chat for anything can motiate businesses to add MCP endpoints to any SaaS service. It will be like extra channel of traffic, leads, orders etc. Some time ago it was popular to have RSS feed on a web site. It was a kind of stanbdard. Then social networks replaced RSS and each web site have buttons to follow a company on social networks. And soon we can see a new thing. Each web site will have a button "Add us to your favorite AI Chat tool". Evarage user will have a list of "MCP servers" in his AI chat tool settings. Similar like modern browsers allow to add extensions and have extensions manager page in settings. ## Possible side effect of this trend When many web sites will really implement a MCP endpoint then we can have next generation of the web - some Web 3.0. Where a data/services of each website is exposed with the standard way the best for reading and understanding by AI or software. And we can be in a situation when business will not need to invest in a web site design because clients will work with a bwe site not with a browser but with a LLM chat interface (text or voice) ##################### # MCP Could Significantly Transform How We Use the Internet Model Context Protocol (MCP) has become a hot topic in discussions around AI and large language models (LLMs). It was introduced to create a standardized way of connecting "external tools" to LLMs, making them more capable and useful. A classic example is the “What’s the weather in…” tool. Previously, each AI chat app had its own custom way of handling that. Now, with MCP, a plugin or integration built for one AI system can work with others too. Recently, we've seen a surge of enthusiasm in building MCP servers for all kinds of services, and this trend is only growing—especially with the use of Server-Sent Events (SSE) as the transport layer. Implementing an MCP server with SSE makes it feel a lot like a SaaS backend designed to serve an LLM or AI chatbot client. There are two main reasons I decided to write this article: - First, it’s widely reported that users are now turning to AI chat apps—especially ChatGPT—instead of Google to look things up. - Second, OpenAI has announced upcoming support for MCP in the ChatGPT Desktop app. They will support both STDIO and SSE transport protocols. Taken together, this points to some interesting changes on the horizon. ## Real-World Use Cases We’ll See Soon ### Sell With ChatGPT (Or Any AI Chat App) Imagine you own a small flower shop and sell bouquets through your website. A customer visits, selects a bouquet, and places an order. Now imagine there's a button on your site: **“Add this store to ChatGPT.”** When clicked, your MCP endpoint is submitted to ChatGPT’s “Add External Tools” screen, and the user approves it. Then, next time the user opens ChatGPT, the interaction could look like this: ``` User: I want to send flowers to my grandma. ChatGPT: The store "Blooming Petals" has some great offers today. I recommend one of these three bouquets: ... Would you like to pick one? User: Option 2 looks good. Send it to my grandma’s address. ChatGPT: Would you like to include a message on the card? User: Yes. You decide. ``` And just like that, the order is placed. **What’s happening in the background:** The flower shop has an MCP server with SSE transport at an endpoint like `https://flowershop.com/mcp`. This server wraps around existing site functionality: search, browse, add to cart, place order, etc. Once the user connects this endpoint to ChatGPT, it becomes a persistent tool in the user’s chat environment. Anytime a relevant topic comes up—like sending flowers—ChatGPT knows it can use your MCP server to help fulfill the request. ### Expose Your Data to AI Chat Here’s another example: imagine you run an online community or forum. Every day, new discussions start and members interact. But now, one of your regular users spends more time in ChatGPT than browsing your site. You can add a button: **“Add this forum to ChatGPT.”** Once they click and confirm, their chat experience becomes much richer: ``` User: What’s new? ChatGPT: There are 5 new discussions in "TechTalk Forum" and one private message. User: What’s the message? ChatGPT: It’s from @alex89. They wrote: “Are you still selling your laptop?” User: Tell them I’m not interested anymore. ``` **What’s happening in the background:** When the user says “What’s new?”, ChatGPT checks all connected MCP tools. It finds that your forum supports the `get_news` and `get_messages` endpoints, uses them, interprets the response, and replies. In this case, your MCP server might expose functions like: - `get_news` - `get_internal_message` - `post_internal_message` With just a bit of development work, any site can integrate directly into ChatGPT (or any MCP-supporting AI chat app), creating an entirely new user experience. ## Voice + LLM + MCP = Magic Now imagine combining this with speech recognition. Some ChatGPT apps already support voice conversations. Add in a few trusted MCP servers and suddenly the experience feels like science fiction: > “Book me a doctor’s appointment next week, and tell my boss I’ll need to leave work early.” And it just… happens. This user has a list of integrated services each with the MCP protocol. Email, Messanger,Calendar. But also preferred Medical company web site, autoservice company web site etc. ## MCP Might Soon Be a “Must-Have” for Websites Just like RSS feeds were once standard, or how websites now include “Follow us on social media” buttons, soon we may see buttons like: > “Add us to your AI chat.” It might become common for users to manage a list of “connected MCP tools” inside their AI chat settings—just like browser extensions today. For businesses, adding MCP support could mean a new source of traffic, leads, or orders. It’s a lightweight way to get integrated into where users spend their time. ## A Possible Side Effect: The Next Web Evolution? If many sites adopt MCP, we could be entering a new phase of the web—perhaps something like Web 3.0 (in the original, data-centric sense). In this new model, websites expose services and data in a standard way, optimized for AI systems. Users may no longer interact with visual interfaces—they’ll interact via chat, text, or voice, powered by LLMs. Businesses might no longer need to invest heavily in traditional UI/UX, because the *chat* becomes the interface. ## Final Thoughts The rise of MCP opens up a new frontier for internet usage. AI chat tools aren’t just about answering questions anymore—they're becoming gateways to services, commerce, and everyday tasks. If you're building a web service or product, adding an MCP interface might soon be as important as having a mobile-friendly design or social media presence. The age of AI-integrated web services is just beginning. Now is the time to prepare.

Previous Post:

Artificial Intelligence
Building MCP SSE Server to integrate LLM with external tools
15 April 2025

In this article, we will build a Model Context Protocol (MCP) server using Server-Sent Events (SSE) to connect a large language model (LLM) with external tools, specifically for executing commands on a Linux instance. We will also implement basic security measures using authorization tokens.

Continue Reading
Building MCP SSE Server to integrate LLM with external tools

Next Post:

Наукова фантастика
Наукова фантастика
5 May 2024

Continue Reading
Наукова фантастика