MCP

What is MCP?

Model Context Protocol (MCP) is an open protocol for connecting AI applications to external tools, resources, and contextual data through a standardized interface.

The underlying problem is straightforward.

Without a common protocol, every AI application may need custom integrations:

AI Application
 ├── Custom GitHub integration
 ├── Custom Database integration
 ├── Custom Slack integration
 ├── Custom Jira integration
 └── Custom Internal API integration

As the number of tools grows, integration complexity grows rapidly.

MCP provides a standardized way for AI applications to discover and interact with capabilities exposed by MCP servers.

A simplified architecture is:

What Problem Does MCP Solve?

MCP primarily addresses interoperability between AI applications and external capabilities.

This can reduce duplicated integration work and make tools easier to expose across compatible AI clients.

MCP Architecture

A conceptual MCP setup contains:

MCP Host

The AI application or environment where the model operates.

Examples can include an AI assistant, IDE, or agent runtime.

MCP Client

The component inside the host that communicates with an MCP server.

MCP Server

A service that exposes capabilities or contextual information to the client.

Conceptually:

The protocol separates the AI application from the implementation details of individual integrations.

MCP Tools, Resources, and Prompts

MCP can expose different types of capabilities.

Tools

Tools represent executable operations.

Examples:

search_repository
create_ticket
query_database
send_message
deploy_service

The model can request a tool operation through the MCP client, subject to the host’s policies and permissions.

Resources

Resources provide contextual information that an AI application can access.

Examples:

documentation
files
database records
configuration
system information

Prompts

Prompts can provide reusable interaction templates or workflows exposed by an MCP server.

The exact behavior and authorization model depend on the host and server implementation.

---

MCP vs RAG vs Agents

These concepts solve different problems and are often used together.

Concept	Primary purpose
LLM	Understand and generate language
RAG	Retrieve relevant external knowledge
Agent	Decide and execute multi-step actions
MCP	Standardize connections between AI applications and external capabilities

A modern AI system may therefore look like:

These are not competing technologies.

They operate at different layers of an AI architecture.