If you’ve not been in the AI hype corners of the internet, we’re in 2025 and MCP’s have taken the AI community by storm. Threadbois hyping it as the next chapter for AI, n8n gurus pivoting to MCP SaaS businesses, and us.. just trying to figure out if people will even adopt the standard long term or not. So what’s different about this one?

New open standards pop up every week like clockwork. None of them really make it to see the light past those 93 views on a random unliked Tweet on X.. let alone serious conversation.. but some do. MCP did. But why?

In a recent article from on Latent Space, Swyx laid out some compelling reasons why the standard hit escape velocity over other standards like Langchain and OpenAPI:

1. MCP is “AI-Native” version of old idea

2. MCP is an “open standard” with a big backer

3. Anthropic has the best developer AI brand

4. MCP based off LSP, an existing successful protocol

5. MCP dogfooded with complete set of 1st party client, servers, tooling, SDKs

6. MCP started with minimal base, but with frequent roadmap updates

Honestly it had to be a company like Anthropic that put it out given OpenAI’s first-mover advantage, but even OpenAI is now backing the standard as of April this year. That said, I’ve gone back and forth with several friends on the topic of MCP’s (and frankly.. confirming what they are and aren’t). MCP’s are.. in fact.. an open standard for interacting with LLM’s and really.. that’s it.

It is a big deal if you are building AI agents and working with LLM’s for work, but I don’t understand the people pivoting what they are doing to build a business around MCP’s. There are a handful that make sense where you might already have marketplaces involving agents, but that’s a small pond to play in. Alternatively, I think enterprise use of them are far more promising in the short term.

Value in AI accrues to the end user

Unlike web2 era tech companies, the party who gains the most is the one who uses AI. Great businesses like Cursor can exist because the multiplicative effect for the user can be MASSIVE. Even capturing a small bit of that can give you the fastest growing company to reach $100M ARR. Insane to think about.

That said it’s a unique time for companies that can string together the API calls to build agents in-house to reap the benefits as the user without having to pay out the nose for an AI agency to build and maintain it for you.. but why MCP?

MCP as the agent gateway

Maybe you’ve seen this image floating around the internet, but this is a visual explainer for the problem that MCP solves. It standardizes how agent and agent applications interact with all of the data and software you need it to touch. The image on the right shows you a case where you might want Cursor to access to everything from Slack, Github, and your local data.. but think bigger for a moment.

In the image we see three API’s that MCP gave Cursor access to.. but what if Cursor wasn’t the only agent in the system? Maybe by some point you have specialized agents that save devs time with code reviews. Hell let’s multiply the agent space to include your call center and emails primarily funneling to a customer service agent. A cold out reach agent making money while you sleep. Another agent vetting potential customers that are already booked for calls throughout the week.

On the other side you have brand assets and guidelines to draw from for your marketing team. CRM system, all social media accounts, email.. I think you get it.

All.. plugged.. into.. MCP.

So for the geeked out engineers out there (like me) building agents for enterprise.. it’s our time.