MCP Explained

MCP stands for Model Context Protocol, and it is a way for different AI components to communicate with each other. Think of it like a universal language that allows different AI tools to work together efficiently.

In AI development, especially when building agentic AI systems, MCP servers are made available to provide necessary functionalities such as email handling, booking services, or data retrieval. These servers can be accessed and used without having to build every component from scratch.

You can find a growing list of MCP servers here: MCP Server List

How MCP Makes AI Development Easier

When building an agentic AI, you often need to provide it with various capabilities such as interacting with a calendar or email system. Instead of building each of these functionalities from the ground up, you can simply connect your agent to MCP servers and use pre-built tools.

For example, suppose you're developing an AI assistant that needs to interact with Gmail or Airbnb. Instead of manually coding all the email-related functions (like sending emails, updating labels, or retrieving messages), you can simply connect to an MCP server that provides all these tools out of the box.

This approach makes development:

• Faster - You don’t have to build out every tool you use from scratch.
• More efficient - Your AI agent can use well-optimized pre-built tools.
• More stable - Since the system is less verbose, there are fewer errors.

The Old Way vs. The New Way

Old Way (Manually Building AI Components)

Previously, if you wanted an AI to handle emails, you had to build an entire Email Agent with functions for sending emails, labeling them, retrieving messages, etc.

Example Process:

1. Build an agent to handle emails.

2. Implement separate functionalities like:

• Sending emails
• Updating labels
• Retrieving messages
• Mark Unread

3. Call this agent from your main application.

Then, you had to call this prebuilt agent from the main AI system. In this example the "Email Agent" is available to be called from the "Tools Agent" along with other pre-built agents.

New Way (Using MCP Servers)

With MCP, you can connect directly to a third-party prebuilt MCP server and call any tool you need without manually building it.

Example Process:

1. Use an MCP Node to connect to an MCP server.

2. List all available tools within the server. This list is automatically pulled from the MCP server when the AI Agent is activated.

3. Use an MCP Node to execute the tools provided by the server.

Now, there’s no need to manually build each agent’s functionality. Everything is pre-built and accessible, saving time and effort.

Advantages of MCP

• Less Development Effort: No need to manually build individual AI sub-agents.
• Easier Maintenance: If a tool’s functionality changes, you don’t have to rebuild the whole system.
• More Scalable: Easily add new features without extensive code modifications.

Example Usage in n8n

In n8n, an automation platform, you can use MCP nodes to quickly integrate powerful AI tools.

1. Add MCP nodes to your workflow.

2. Instantly gain access to several new AI functionalities without additional coding, for example the "MCP Client" or "MCP Client Tool".

Conclusion

MCP is revolutionizing the way we build AI by making pre-built tools available through a simple, standardized protocol. Instead of spending countless hours manually coding functionalities, developers can now leverage existing AI capabilities and focus on building smarter, more efficient systems.

If you’re working on agentic AI, consider using MCP to save time, reduce complexity, and improve system stability!

Paul Hankin is the author of:

AI Adoption: A Practical Guide for Business

and

AI and Law: Navigating the Future