MCP: Celebrating a Year in NLP

The Model Context Protocol (MCP) is turning one. Since Anthropic publication introducing the MCP on November 25, 2024, it has become the missing piece allowing AI models to act upon our development environment using natural language, without relying on proprietary integrations.

MCP has given shape to a simple idea: if the LLM is the brain, then it needs hands to execute real actions. And those hands are the tools that MCP connects.

What is MCP, Explained Straightforwardly

MCP is an open protocol that enables an AI model to use external tools: commands, APIs, services, databases, validators, linters… all through natural language and in a standardized way.

Instead of every company inventing its own system, MCP establishes a common language between:

MCP Servers → where the tools live.
MCP Clients → the LLMs and IDEs that use them.

If you install an MCP server, any compatible client will be able to interact with it in your own language.

- Create a repository on my GitHub account called "my-mega-project" 
- Give me the list of repositories for the organization "my-company"
- Give me the list of issues for the repository "my-repo"
- Create an issue in the repository "my-repo" with the title "my-issue"

And you have plenty to choose from. Check out this catalog of MCP tools to see those that Copilot considers might be of interest to you. It's not all of them, but there are many more.

Why It Matters

Before MCP, every editor or provider had its own mechanism for connecting AI to the environment. The result: fragmentation, duplication, and limitations.

With MCP:

The LLM stops being just text and becomes an actor in the system.
You can request tasks in natural language that previously required scripts or manual steps.
All IDEs that support it can use the same tools.
Companies can create their own catalog of internal capabilities.

It is a standard you can know and use in any editor or tool that supports it. This is its official website.

The Practical Metaphor

Think of AI-assisted development like this:

The LLM is the brain: it thinks, plans, suggests.
MCP acts as the nerves: it transmits instructions.
The tools are the hands: they execute real actions.

Without MCP, the "brain" can only talk. With MCP, it can act.

Impact in 2025

This year we have seen three clear effects:

1. IDEs Become Tool-Aware

It doesn't matter if you use VSCode, Cursor, or any compatible editor. MCP allows the AI to use real tools in your project, from tests to internal APIs.

2. Agents Do More Than Write Code

They can now execute commands, analyze repositories, consult internal services, and validate business rules.

3. Growing Ecosystem

There are already dozens of ready-to-use MCP servers: GitHub, databases, linters, CI/CD tools, and third-party APIs. And anyone can create their own.

And the important thing: anyone can create their own MCP server. Here is an accessible tutorial: Building your first MCP server: a beginners tutorial

This has opened the door to integrating AI into internal systems without going through proprietary tools.

Conclusion

MCP has turned one, and although at first it seemed like just another technical proposal, today it is clear that it has changed the architecture of AI tools for development.

It did something the industry needed: standardizing how a model can act upon the developer's environment.

If 2023 was the year of the models, and 2024 was the year of the AI editors, 2025 is shaping up to be the year of connected agents. MCP is the foundation upon which this new layer is built.

Useful Links

Original Presentation (2024): https://www.anthropic.com/news/model-context-protocol
Official Documentation: https://modelcontextprotocol.io/docs/getting-started/intro
MCP Tools Catalog: https://github.com/mcp
Tutorial for creating an MCP server: https://dev.to/debs_obrien/building-your-first-mcp-server-a-beginners-tutorial-5fag

MCP: One Year of Action in Natural Language