Model Context Protocol (MCP)
MCP is a standard protocol for connecting AI applications to external tools and data sources, so each new integration doesn’t need custom, one-off code.
Prerequisites
Why a Protocol?
Before a standard existed, connecting an AI application to a new tool or data source usually meant writing custom integration code for that specific combination. Every new application, and every new tool, multiplied the number of one-off integrations needed. MCP defines a standard way for applications and tools to talk to each other, so a tool built to speak MCP can work with any MCP-compatible application, and vice versa — without custom glue code for every pair.
How It Fits Together
AI Application
Wants IntegrationNeeds a tool or data source it lacks itself.
MCP Client
Standard InterfaceNo custom glue code for this specific pairing.
MCP Server
Any Compatible ServerWorks with any MCP-compatible application.
Tools & Resources
The Actual IntegrationWhat the application ultimately gets access to.
- MCP Client
- The application or agent side that connects to MCP servers to discover and use what they offer.
- MCP Server
- Exposes a set of tools, resources, or data over the protocol, to any client that connects.
- MCP Tools
- The callable actions an MCP server exposes — conceptually the same idea as the tools in tool calling, offered through a standard interface.
- MCP Resources
- Data or content an MCP server exposes for a client to read as context, rather than as a callable action.
How MCP Relates to Other Concepts
MCP is not a competing idea to function calling, tool calling, or agents — it standardizes the connection between them and the outside world.
- Function calling / tool calling — the mechanism by which a model requests and uses a capability. MCP standardizes how that capability is exposed and discovered, rather than replacing the mechanism itself.
- APIs — an MCP server is often a wrapper around one or more existing APIs, exposed in a way an MCP client already knows how to talk to.
- Agents — an agent can use MCP as its way of discovering and calling tools, instead of every agent needing custom integration code per tool.
Common Mistakes
Confusing MCP with an LLM
MCP is a protocol for connecting applications to tools and data — it doesn’t generate text or reasoning on its own.
Confusing the protocol with an agent
MCP standardizes how an application discovers and calls tools; the reasoning loop that decides what to do is a separate concern, typically handled by an agent.
Assuming MCP automatically makes tools safe
Standardizing the connection doesn’t remove the need for authentication, permissions, and validation on what a tool is allowed to do.
Interview Question
What is MCP, and what problem does it solve?
MCP, the Model Context Protocol, is a standard for connecting AI applications to external tools and data sources. Before a shared standard, every application needed custom integration code for every tool it wanted to use, which multiplies quickly as both sides grow. MCP defines a common client-server interface — an MCP client in the application connects to MCP servers that expose tools and resources — so a tool built once can work with any compatible application. It doesn't replace function or tool calling; it standardizes how those capabilities are discovered and connected.
What an interviewer may ask next
- How does MCP relate to function calling and tool calling?
- What is the difference between an MCP tool and an MCP resource?
- Does using MCP automatically make a tool safe to call? Why or why not?
Explain It in 30 Seconds
MCP is a standard protocol for connecting AI applications to external tools and data sources, so a tool built once can work with any MCP-compatible application instead of needing custom integration code for every pairing. An MCP client, on the application side, connects to MCP servers that expose tools and resources. It doesn’t replace function calling or agents — it standardizes the connection those patterns rely on.