Beyond chatbots: how model context protocol (MCP) enables LLMs to take action

In the world of AI and LLMs, there is a growing need to interact with systems. This goes beyond using a chatbot to respond to a prompt; it should be possible to perform actions using a chatbot, such as updating the status of an invoice, ordering an item on your e-commerce, and so forth.

By their nature, large language models (LLMs) are used to respond to prompts using natural language. They are not designed to interact with your systems to execute actions. To be able to bridge the gap and cover these types of use cases, the Model Context Protocol (MCP) was recently introduced.

While this is a rather new protocol, it is gaining a lot of traction and we’re seeing usage explode amongst both vendors offering the technology to support this protocol, or companies offering their services through an MCP server.

What is MCP?

The Model Context Protocol (or MCP) is an open protocol that standardizes how applications provide context to LLMs. It is often referred to as a USB-C port for AI applications. Through MCP, LLMs can interact with your systems and retrieve relevant information or execute actions. The advantage of MCP is that is a standard that can be used for all types of interfaces and systems such as APIs, legacy applications, file systems, databases, etc.

As a side note: if you are familiar with Retrieval Augmented Generation (RAG), these two concepts sound very similar, but there are some key differences and reasons to choose one or the other. RAG and MCP can both be used to access data from an outside source, however, MCP can also be used by an LLM to perform actions.

RAG is most suited for AI search, where the latest and most relevant information needs to be retrieved, whereas MCP is most suited for agentic AI use cases, where actions need to be initiated.

How does MCP work?

MCP (Model Context Protocol) works in a client-server architecture with the following components:

• MCP hosts: Programs that want to access data or functionality and are powered by an LLM.
• MCP clients: A client that maintains a 1:1 connection between the host and an MCP server. Each server requires a single client.
• MCP servers: Application that exposes functionality from an application or data source through the Model Context Protocol.
• Data sources and services: Applications that MCP Servers interact with and that contain the data and services that can be leveraged by the LLM.

What is an example of MCP?

Below are two examples of MCP:

Perform actions

Q: Update the last order of ACME and set the status to processed.
A: The status of the last order of ACME with number 1354643 has been set to processed.

Retrieve relevant data from data sources

Without MCP
Q: I want to retrieve the last transaction on my checking account.
A: Use your banking app for this information.

With MCP
Q: I want to retrieve the last transaction on my checking account.
A: Your last transaction was 250USD at Amazon.

How can you use Fusion to tackle MCP use cases?

Axway Amplify Fusion is a low-code or no-code integration platform that solves simple to complex enterprise integration use cases and patterns.

Primary use cases are application integration, data integration, and APIfication (getting apps to talk to each other via application programming interfaces, or APIs). The solution is offered as a pure software-as-service (SaaS), hybrid, or customer-managed deployment.

Specifically, Fusion provides the essential building blocks for implementing MCP by:

1. Providing an integration layer between your applications and the MCP server
2. Delegating authentication to APIs and other applications
3. Functioning as an MCP Server
   • Use the Integrated MCP server or
   • Use your own MCP server
4. Automatically generating the MCP Server interface when using the integrated MCP server
5. Supporting any protocol: API, Legacy, DB, Files, and more
6. Offering a marketplace for MCP servers via integration with Amplify Engage
7. Implementing rate limiting

The following shows an example diagram of an MCP implementation using Fusion.

Amplify Fusion is a critical component in helping companies introduce MCP, which is essential to help LLMs perform actions and enable agentic AI. This blog has shown the principles behind MCP, the value it can bring to your business, and how Fusion can help you implement an MCP scenario.

In my next blog, I’ll share step-by-step instructions for setting up an MCP server on top of your existing APIs with Amplify Fusion.