Practical MCP with FastMCP & LangChain

Retrieval Augmented Generation vs Model Context Protocol Resources

It is easy to confuse RAG systems and MCP resources because both deal with giving extra information to a model. But they are not the same thing, and they solve different problems.

Retrieval-Augmented Generation, or RAG, is a way of improving a model’s answers. The idea is simple: before asking the model to respond, you search for relevant information and include that information in the prompt. This search step might use embeddings, vector databases, similarity search, or ranking systems.

Here is a concrete example:

Imagine you build a chatbot for a company that sells laptops. A user asks: How do I reset the XPro 15 BIOS?.

The model does not know your company’s specific instructions. So instead of answering immediately, your system searches your internal documentation. It finds a section in the support manual that explains the reset steps. That relevant text is then added to the prompt along with the user’s question.

The model reads both the question and the retrieved instructions, then generates a clear answer based on that context.

So RAG works in two steps: