MCP – Will Managers Replace Developers?

Model Context Protocol (MCP) is quietly becoming one of the most transformative technologies of 2025. Introduced just this March, MCP has already been embraced by the majority of major AI platforms – and an open-source ecosystem has sprung up around it at breakneck speed.

But what exactly is MCP? And why might it radically shift the relationship between business users and developers?

🧩 What Is MCP?

MCP stands for Model Context Protocol – a standardized, language-based protocol that connects:

• Different LLMs (large language models)

• Your internal data sources (APIs, databases, documents)

• Specialized task-oriented services (called agents)

Think of MCP as the glue that binds together disparate services and makes LLM-driven automation practical, scalable, and extensible. It's the next logical step after OpenAI-style function calling.

🕸️ Agents Everywhere – An AI-Powered Service Mesh

At the heart of MCP is the concept of agents – small, independent services that perform narrowly defined tasks. These agents:

• May be AI-driven or traditional

• Are controlled and orchestrated by an LLM

• Use natural language as the interface

• Pass structured data between each other like a pipeline

Let's look at a real-world example of what's now possible.

Imagine you want to reach out to existing customers with a product upgrade offer. Your agents might include:

• 🔎 Reading CRM data via API

• 📄 Parsing product information

• ✍️ Composing personalized emails

• ✉️ Sending emails

• 🌍 Translating messages

You – as a business user – write:

"Generate a sales email for customers who purchased from product line X. Suggest a relevant upgrade or add-on. Offer a 5% discount to customers who spent over $1000 last year. Send drafts to me for review."

Within moments, you'll receive a set of customized emails. You can adjust them via simple prompt changes. Once satisfied, say:

"Send to selected customers."

And that's it. Time for golf. 🏌️

⛔ No More IT Bottlenecks

This is the real disruption.

The traditional flow of "manager writes specs → developer implements → repeat" is being bypassed.

With MCP:

• The IT team sets up the environment and agents

• LLMs handle orchestration

• Managers and other non-technical users drive the logic and flow with plain language

This eliminates misunderstandings, development delays, and context loss between departments. The IT crowd becomes the infrastructure team – not a bottleneck.

And the tooling is arriving fast.

🛠️ Check out: https://github.com/modelcontextprotocol/servers
You'll find connectors and templates for nearly every enterprise tool in use today – from PostgreSQL to Notion to LangChain. And this happened in less than six months.

📈 Rethinking the Role of Managers

The most profound change won't be technical – it'll be organizational.

As MCP gains traction:

• Managers will need new skills: automation logic, prompt engineering, agent design

• Business processes will become modular and language-driven

• The speed of iteration and experimentation will explode

One early adopter worth watching is Czech tech unicorn Rohlík. Its CEO, Tomáš Čupr, was among the first to embrace AI-driven business workflows and even shares his AI tools publicly on GitHub. 🧠💡

💡 What Comes Next?

MCP represents a revolutionary shift in how we design systems, interact with data, and automate workflows. It blurs the lines between user and developer, spec and execution, tool and team member.

In upcoming posts, I'll dive deeper into:

• How to build your own agents

• Best practices for orchestrating workflows with LLMs

• Governance, security, and auditability in the MCP world

Stay tuned – this is just the beginning.

🧠 Thoughts, feedback, or war stories? Drop me a mail or message!