Interesting thoughts regarding MCPs being the future App Store/Platform. I don't know that I agree but I don't necessarily disagree either. Time will certainly tell.

To me, MCP feels more like an implementation detail, not something that most people would ever use directly. I would expect that the future would be some app distributed through existing channels, which bundles the MCP client into it, then uses a server-side component (run by the vendor of course) to get the real work done. As much as I like the idea of people installing the servers locally, that future seems like a Linux nerd/self hosted type of activity. I just can't imagine a typical mac or windows non-power-user installing one directly. Just the idea that they would need to install "two apps" is enough to confuse them immensely. It's possible some might bundle the server too and run it locally as needed, but even in that case I think MCP is completely invisible to the user.

  - talks to cloud LLM
  - talks to local MCP servers

  - connects to company cloud (this lets the company collect usage data)
  - forwards tasks to the cloud

  - connects to company cloud to get a users tasks
  - executes the tasks (this lets the company use the users hardware, saving cloud costs)

> Think of MCPs as standardized APIs—connectors between external data sources or applications and large language models (LLMs) like ChatGPT or Claude.

This is incorrect.

MCP is Model Context Protocol.

You didn't "build an MCP", you implemented an MCP server. Lighttpd is not "an HTTP", it's an HTTP server. wget is also not "an HTTP", it's an HTTP client. Lighttpd and wget are different enough that it's useful to make that distinction clear when labeling them.

dnsmasq is not "a DHCP", it's a DHCP server.

This distinction also matters because it is certain that we will see further protocol iterations so we will indeed have multiple different MCPs that may or may not be compatible.

MCP might be one of the few technology pieces where more articles have been written about it than the actual use-cases being built.

It is like the ERC20 era all over again.

My prediction: there will be no standard protocol, clients will do whatever works for them, and devs will do whatever it takes to be installable on those clients. Just like mobile.

If LLMs are so smart, why do they need a custom "MCP" format to what's commonly known as a normal API? Why can't they just call normal APIs?

Extending this thought: why would there be any difference between offering data behind an API, and offering data behind a "MCP api"? At the end of the day, the underlying data will be the same (weather, stock market info, logs, whatever), it seems LLMs just needs this to be "standardized", otherwise it doesn't get it (?).

Furthermore..! LLMs can already crawl web pages just fine using (true) restful technologies. So why would there be need for other, new, special APIs when it's enough to expose the same data on a normal website?

I don't get it.

>>MCP Affiliate Shopping Engines

As someone else once said, I want a Grocery Shopping Engine. "Here's my shopping list, taking into consideration delivery times and costs, please buy this for the lowest cost from any combination of supermarkets and deliver by day after tomorrow at the latest."

If MCPs gave the LLMs a window into all the major supermarkets home shopping sites that looks like it's a step closer.

MCP is perhaps the biggest attack vector I've seen people willingly adopt simply for FOMO. Nothing about implementing it is defined or tractable. Even logging its use is extremely complicated.

MCP in this context means "Model Context Protocol"

I thought it might be "managed cloud providers", but perhaps I'm too optimistic for a change

If you're sold on MCP, what was your "wow" moment? I've read the docs and tinkered a bit but it was a decidedly "meh" experience personally. It seems very similar to ChatGPT Plugins, and that was a flop. I don't really like the fuzzy nature of the architecture, where I never know what server will be invoked. And I have to manually opt-in to each server I want to use? To be unexpectedly useful, it seems like I would have to opt-in to tens or hundreds of servers. Yet I've heard that clients start to struggle once you have more than 20 servers plugged in...? Please excuse any fundamental errors I've repeated here, if any...

Anyone else explored a plugin approach to bundle tools with the client? I started down that path long before MCP was a thing. I do plan to add MCP and A2A support at some point. But for my program it's generally not necessary. You just go to the admin and install plugins with the tools. https://github.com/runvnc/mindroot

How does discovery work with MCP? Is there a way to make it fairly seamless?

Are there any open-source MCP 'app stores' currently available? I am considering building one for my employer. A place to host internally built MCPs. It would be useful to know if there is something already out there.

My experience so far with MCP is also that this is not ready yet. There are a few working use cases, but generally this stuff is pretty raw.

we’ve been building most of what OP has written about with https://mcp.run

We started doing this the day Anthropic released MCP in November last year. Our company has always been devoted to secure plug-in system technology having built Extism, a WebAssembly plugin framework.

We immediately saw MCP as the plugin system for AI and knew it would be significant, but were concerned about the security implications of running MCP servers from untrusted parties and using the STDIO transport which makes user systems vulnerable in ways we weren’t ok with.

So we built mcp.run which is a secure implementation of the protocol, running servers in fully isolated & portable wasm modules. They must be allow-listed to access files & network hosts, and cannot access any part of your system without your explicit permission.

They also run everywhere. Each server (we call them servlets) on mcp.run is automatically available via SSE (soon HTTP streaming) as well as STDIO, but can also be embedded directly into your AI apps, no transport needed, and can run natively on mobile!

We are excited about MCP and glad so many are too - but we really need more security-oriented implementations before it’s too late and someone suffers a seriously bad exploit - which could tarnish the ecosystem for everyone.

The Master Control Program has no future!

The future? MCP is fighting even to have a present.

MCPs tries too hard to signal XHR for AI, but nobody wants to earnestly deal with the consequences of AI interfacing in a wider context of mis-/disinformation, hallucination and generally letting it talk to stuff in a semi-unprompted manner.

I’d just like to interject for a moment. What you’re refering to as Linux, is in fact, GNU/Linux, or as I’ve recently taken to calling it, GNU plus Linux. Linux is not an operating system unto itself, but rather another free component of a fully functioning GNU system made useful by the GNU corelibs, shell utilities and vital system components comprising a full OS as defined by POSIX.

Many computer users run a modified version of the GNU system every day, without realizing it. Through a peculiar turn of events, the version of GNU which is widely used today is often called Linux, and many of its users are not aware that it is basically the GNU system, developed by the GNU Project.

There really is a Linux, and these people are using it, but it is just a part of the system they use. Linux is the kernel: the program in the system that allocates the machine’s resources to the other programs that you run. The kernel is an essential part of an operating system, but useless by itself; it can only function in the context of a complete operating system. Linux is normally used in combination with the GNU operating system: the whole system is basically GNU with Linux added, or GNU/Linux. All the so-called Linux distributions are really distributions of GNU/Linux!

> You didn't "build an MCP"

The author explicitly states he built 2 MCP servers, not 2 MCPs, so I don’t know where your beef is coming from

Purists perpetually decry the zeitgeist's sloppy terminology.

Words that climb the Zipf curve get squeezed for maximum compression, even at the cost of technical correctness. Entropy > pedantry. Resisting it only Streisands the shorthand.

“The map is not the territory.”

Do you mean "notifications", ie a core feature of every computer and phone?

I also did not get it, but now I get it a bit, I think.

Look at it this way. You have to get some work done - maybe book a flight ticket. So you go to two sites - first you go to flight fare comparison, then you book the ticket on the airline website. And you have to do it in code.

There are two ways you can do it.

First Way 1. Understand the API of the flight comparison portal. 2. Understand the API for the airline website. 3. Write code which combines both these API and does the task.

Second Way 1. Message a coder friend who knows the API of the flight comparison portal and ask him to write code to get the cheapest flight. 2. Message another coder friend who knows the API of the airline portal and ask him to book a flight.

Both ways are possible, but which one do you think is Less Work ? Which one is 'cognitively' easier ? Which one can you do while driving a car with one hand ?

It should be clear that the second way is easier. Not only is the second way easier, but if the task requires multiple providers and a lot of context, it might be the only way possible.

The first way is analogous to LLM's doing API calls. The second way is analogous to LLM's doing MCP Servers. MCP servers reduce the cognitive cost to do a task to the LLM - which dramatically increases their power.

> If LLMs are so smart, why do they need a custom "MCP" format to what's commonly known as a normal API? Why can't they just call normal APIs?

LLM's _can't_ just call APIs, because all they can do is generate text. The LLM can _ask_ you to run some code, but it has no ability to run code directly. MCPs are basically a way for LLMs to signal intent to make an API call, along with a list of white listed APIs, and documentation for using them, and preloaded credentials with whatever permissions you want to give them.

I don't understand why just using something like openapi specs didn't become the "normal" thing to do. We already have APIs for pretty much everything, why do we need a new protocol that wraps around an existing api?

Ye its also funny to me. On the one side people are saying: Look we have computer use, browser use etc. so we don't need an api! And on the other side saying, look apis are way too complicated we need our own protocol!

They actually can. I have found myself getting more use out of a terminal MCP and providing OpenAPI specs than bespoke MCP servers.

Bespoke MCP's right now are a convenience.

Well, LLMs are not so smart for starters.

The first use case I found relevant and useful was the Supabase MCP server, allowing Cursor's agent to query my Supabase project. It meant no longer describing my database to Cursor, it could simply go and get the information it needed, as needed.

Wrote an MCP to hook into my logging so I could get Claude + Cursor to quickly answer "hey, why did request 20394 from yesterday evening fuck up?". It goes into the logs and finds the exception, hunts down the line and then tells me whats up. Of course, left unchecked it tries to fix the problem too but I've spent countless lines of prompt engineering to have it never attempt to "just start writing code".

My "wow" moment was when I wrote an internal MCP server so that Claude Code could access our test databases. It was a tiny amount of code, simple to connect up, and immediately gave Claude Code a way to directly validate queries. It's been useful in numerous scenarios since then and got me thinking about additional MCP-based tools it might be nice to have.

For me it was implementing a simple `execute_terminal_command` tool along with hooking up to my company's jira and gitlab (dont worry security gurus, for the command line, I have a hardcoded list of allowed read-only commands that the LLM can execute, and both jira and gitlab servers likewise have readonly options.)

What I will say is I agree there should be an option to get rid of the chat confirmations of every single new tool call in a chat - as well as building a set of "profiles" of different tools depending what I'm working on. Also strongly agree there needs to be an internal prompt possibility to explicitely tell the LLM what tool(s) to favor and how to use them (even in addition to the descriptions / schemas of the tools themselves) I opened an issue on the anthropic repo exactly about this: https://github.com/modelcontextprotocol/typescript-sdk/issue...

Writing an internal MCP server to link our API layer to Augment/VSCode so that our Frontend developers can ask in plaintalk about API details. With over 1000 endpoints, it lets the devs find the endpoint, and more importantly the GQL fields, quickly. After some dogfooding we plan to open it up to our clients as well.

We have a software templating tool we use internally to start new projects. I wired an MCP server into it and now I can just ask cursor to start a new project for me, and it'll go through our list of templates, find the most useful one, create a new project, read the documentation for it, and then be ready to building it and adding new features right away.

Yep. It is currently a Meh experience as said in the OP because the UX sucks. The idea is take a step back and imagine what could it become if those are fixed.

Btw, one of my favorite MCPs is a Whois MCP so I can ask Claude Desktop to brainstorm domain names and then immediately check if they are available :).

It’s clunky but I am still using it :)