MCP Explained - Why Everyone Is Talking About This AI Evolution

MCPs Explained: Why Everyone's Talking About This AI Evolution

The recent buzz around MCPs has left many wondering what exactly they are and why they matter in the AI landscape. In this post, I'll break down Professor Ross Mike's excellent explanation of MCPs, their significance, and the potential startup opportunities they present.

 What Are MCPs?

MCP stands for Message Composition Protocol. In essence, it's a standardized way for Large Language Models (LLMs) to communicate with external tools and services.

To understand MCPs, we first need to understand how LLMs have evolved:

 Stage 1: Basic LLMs

In their most basic form, LLMs like early ChatGPT versions could only predict text and respond to queries. They couldn't perform meaningful actions like sending emails or accessing external data. They were limited to generating text based on the data they were trained on.

As Professor Mike puts it: "LLMs by themselves are incapable of doing anything meaningful... the only thing an LLM in its current state is good at is predicting the next text."

 Stage 2: LLMs + Tools

The next evolution connected LLMs to external tools and APIs. This allowed AI assistants to search the internet, access databases, and perform specific tasks. Services like Perplexity and Brave Search exemplify this approach.

However, this approach has significant limitations:

- Connecting multiple tools becomes cumbersome

- Each tool speaks its own "language" (different API structures)

- Changes to one service can break the entire system

- Creating a cohesive experience across tools is difficult

Stage 3: MCPs

MCPs represent the next step by providing a standardized layer between LLMs and external services. Think of MCPs as a universal translator that allows the LLM to communicate seamlessly with all connected tools.

 The MCP Ecosystem

The MCP ecosystem consists of four main components:

1. **MCP Client**: The LLM-facing interface (e.g., Tempo, Windsurf, Cursor)

2. **MCP Protocol**: The standardized communication method

3. **MCP Server**: Translates between the protocol and external services

4. **Service**: The actual tool or API providing functionality

What's clever about Anthropic's approach is that they've put the responsibility of creating MCP servers on the service providers. This means companies that want their services to be accessible to LLMs now have a clear standard to follow.

Why MCPs Matter

MCPs solve several critical problems:

1. **Standardization**: Instead of "gluing" different tools together, each with their own implementation, MCPs provide a unified approach.

2. **Resilience**: When a service updates its API, it's the service provider's responsibility to maintain MCP compatibility, reducing breakage.

3. **Capability**: MCPs enable more complex, multi-step workflows that were previously difficult to implement.

4. **Developer Experience**: Creating AI assistants that can use multiple tools becomes significantly easier.

As Professor Mike explains: "What MCP does is unify the LLM and the service. It creates this layer where the service and the LLM can communicate efficiently."

 Challenges with MCPs

It's not all "sunshine and rainbows" yet. There are still technical challenges:

- Setting up MCP servers can be cumbersome

- The standard is still evolving

- There are local setup complications to overcome

However, once these kinks are worked out, we'll enter a world where LLMs become dramatically more capable through standardized tool access.

Startup Opportunities

When asked about business opportunities around MCPs, Professor Mike offered these insights:

1. **For Technical Founders**: Consider building infrastructure that makes MCP implementation easier. One idea he shared was an "MCP App Store" where people could browse, install, and deploy MCP servers with a simple click.

2. **For Non-Technical Founders**: Stay closely attuned to which platforms are building MCP capabilities. When standards solidify, there will be opportunities to create integrations and experiences that weren't previously possible.

The key is to watch the space closely. As with other protocol innovations like HTTPS or SMTP, once standards are established, new business opportunities will emerge.

 Conclusion

MCPs represent a significant step forward in making AI assistants more capable and reliable. While we're still in the early days, understanding this evolution puts you ahead of the curve when it comes to identifying future opportunities.