Nothing so transformational I would argue but at the same time it’s a nice to have that makes developing AI solutions easier. A lot of people use the analog of a USB C connector so in the same way that USB C is not truly transformational, it does not enable you to do more things, but its a nice to have, since it reduces the worry and the occasional pain of mismatch ports.

To bring the conversation back to AI models, you could very well connect AI models with tools before MCP. In fact I would argue that using those tools effectively was not limited by a fragmented way to interact with those tools but rather from the AI model inability to use the right tool at the right time. Remember ChatGPT plugins?

What about the MxN integration problem? If you have M AI models and N tools, you need to build MxN connections, isn’t MCP solving that? It does, but how often do you have M AI models and N tools? AI providers definitely have that problem cause they want to compare their model to the competition but when developing a solution, you most likely pick a reasonable choice, be it Claude, GPT, Gemini etc, connect it with the tools you need, and iterate to make it work as per your requirements.

To get the story straight though, I am not saying MCP is only hype, the same way that USB C is not only hype. It is definitely a very useful standard that makes developing AI solutions much easier, I just don’t think its that it enables you to build more complicated solutions that you could before 👌