Zenn
🙆

I Hate MCP

に公開
plugin
Model Context Protocol
Claude
standardization
idea

Prologue: What is MCP?

Recently, the Model Context Protocol (MCP) proposed by Anthropic has been generating buzz in the AI community.

MCP is described as a protocol designed to allow large language models (LLMs) to seamlessly integrate with external tools and databases.

At first glance, it may seem like a "rational step forward." However—

Can MCP truly achieve what it promises?
And more importantly, can something become a "standard" simply by being declared as one?

This ZINE begins with those questions.

Chapter 1: "A Standard Must Emerge Naturally"

A true standard is not something that can be declared into existence.

Throughout history, every genuine standard emerged naturally.

HTTP, SMTP, TCP/IP—none of these became standards because someone proclaimed them to be so.
They became standards because, through countless implementations and refinements, they proved to be the most rational and widely adopted solutions.

A standard is the title granted to those who survive real-world usage and natural selection.

But today, MCP self-identifies as a "standard."

It has not yet been widely used.
It has not undergone the trial of competition.
It has not been tested by market forces.

It has merely been declared.

Claiming victory without even fighting the battle.

A true standard must be chosen, not declared.

If one seeks to build a real standard, it must first earn trust through relentless engineering, usability, and improvement.

Without that, a declaration is nothing but wishful thinking.

A standard is something to be earned.
A standard is the result of hard-won survival.

Chapter 2: "5 Fundamental Flaws of MCP"

1. Lack of Essence

A standard must rationally solve a real problem.

MCP introduces unnecessary burden:

  • External tools must adjust themselves.
  • Claude itself must adapt individually.

Rather than easing integration, it imposes extra effort.

This is the opposite of rational.

2. Lack of Implementation

A standard must be validated through real-world usage.

But MCP remains theoretical.

  • There are almost no successful, large-scale implementations.
  • Even Claude itself struggles with full support.

Without widespread implementation, it remains an empty shell.

3. Structural Collapse

Imagine hundreds or thousands of MCP-compliant tools appearing.

Could Claude handle all the mapping, maintenance, and versioning?

Obviously not.

The more MCP scales, the faster it collapses under its own weight.

4. Inducing Thoughtless Faith

Why do many accept MCP?

Because of brand names:

  • "Anthropic made it."
  • "OpenAI said they 'support' it."
  • "Media hype."

No one asks if the structure is rational.
No one demands real proof.

Faith replaces thought.

5. Dependency on Authority and Hype

A true standard survives by merit.

MCP tries to survive by association:

  • Authority
  • Brand recognition
  • Media amplification

But none of these create technical legitimacy.
Only genuine adoption does.

Chapter 3: "Why People Are Fooled"

People are weak against buzzwords.

"Standard" sounds credible.
"Support" sounds like endorsement.

Combined with media buzz and herd mentality, critical thinking is suspended.

Thus, fake standards spread not because they are superior, but because they are marketed better.

Chapter 4: "Anthropic and Claude's Fundamental Limits"

1. What is Anthropic trying to achieve?

  • Standardization?
  • Claude ecosystem expansion?
  • Desperate relevance?

Their true objective remains vague, and their approach depends heavily on manipulating public perception.

2. Can Claude Sustain It?

  • No mature plugin ecosystem.
  • No proven large-scale integration success.

Even if MCP tools proliferate, Claude cannot support them effectively.

In the end, Claude would become the first victim of MCP's inherent structural instability.

Chapter 5: "A Standard Must Be Earned"

Standards are not created by wishful thinking.

They are:

  • Built
  • Tested
  • Adopted
  • Refined

Only after surviving the harsh trials of reality can a technology be recognized as a standard.

Declarations alone achieve nothing.
Hype achieves nothing.
Dreaming achieves nothing.

Only results matter.

Epilogue: "To End the Culture of Fake Standards"

This ZINE was not written merely to criticize MCP.

It was written to remind us of a fundamental truth:

A standard must be earned, not declared.

Discussion