# What is MCP in Claude Code (Model Context Protocol)

> Understand what the Model Context Protocol is, how it connects Claude Code to external tools, and why it opens up a world of new possibilities.

- Source: https://doutorai.pt/en/blog/what-is-mcp-in-claude-code
- Author: Miguel Diamond (https://doutorai.pt/en/author/miguel)
- Category: Claude Code
- Published: Jun 29, 2026
- Site: Doutor AI (https://doutorai.pt/en) - AI tools for small businesses

## Quick answer
MCP (Model Context Protocol) is an open protocol that lets Claude Code connect to external tools and data sources, such as databases, APIs, file systems and web services. Instead of the assistant working only with text, MCP gives it the ability to interact with the real world.

- Open protocol to connect Claude Code to the outside world
- Connects to databases, APIs, files and web services
- Lets Claude Code take real actions, not just generate text
- Extensible: any tool can have an MCP server

## What is MCP

The Model Context Protocol, or MCP, is an open protocol developed by Anthropic that defines a standard way for Claude Code to connect to external tools and data sources. Without MCP, Claude Code essentially works with text: it reads files, writes code, answers questions. With MCP, it can interact with databases, query APIs in real time, access remote file systems and even control external applications.

The word "protocol" may sound technical, but the idea is simple: MCP defines a common language between Claude Code and the tools you want to use. Any tool that implements that language can be used by the assistant. To better understand the context where MCP fits, visit our article on [what Claude Code is](https://doutorai.pt/en/blog/what-claude-code-is-and-what-its-for) and the [main Claude Code page](https://doutorai.pt/en/claude-code).

## How it works in practice

MCP works with two main components:

- **MCP client**: this is Claude Code. It makes requests to external tools and uses the data it receives to carry out your instructions.
- **MCP server**: the external tool (a database, GitHub, a web service, etc.) that receives requests from Claude Code, performs actions and returns data.

When you configure an MCP server, Claude Code gains access to that tool. For example, with an MCP server for GitHub, the assistant can browse repositories, create issues or view pull requests directly within your work session, without needing to leave the terminal or copy and paste information manually.

## Examples of MCP servers

The MCP ecosystem is growing fast. Here are some examples of what you can already connect to Claude Code:
MCP serverWhat it lets you doGitHub MCPRead repositories, create issues, view pull requestsPostgreSQL MCPQuery and write to databases directlyGoogle Drive MCPAccess and edit documents in Google DriveWeb Search MCPSearch the internet in real timeFile System MCPAccess files and folders on your local systemSlack MCPRead and send messages in Slack
The list keeps growing as more companies and developers create MCP servers for their tools.

## MCP vs plugins: what's the difference

You may be familiar with the concept of plugins in other AI assistants. MCP is conceptually similar, but with important differences:

- **Plugins** are often proprietary and specific to a single assistant. To work with a different assistant, you need to rewrite the plugin.
- **MCP** is an open, standardized protocol. An MCP server written for Claude Code can potentially work with any assistant that supports the protocol, reducing duplicated effort for anyone building integrations.

In practical terms, MCP tends to be more flexible and to allow deeper integrations than traditional plugin systems.

## How to get started with MCP

If you are using Claude Code and want to try MCP, the easiest path is to start with an existing MCP server for a tool you already use day to day. Anthropic and the community maintain lists of MCP servers available for the most common tools.

The general process is: install the MCP server for the tool you want to use, configure the required permissions and add the server to the Claude Code configuration. From then on, the assistant has access to the tool and can use it in your work sessions without manual intervention from you.

To deepen your knowledge of Claude Code's capabilities, also check out the articles on [skills](https://doutorai.pt/en/blog/claude-code-skills) and [agents](https://doutorai.pt/en/blog/agents-in-claude-code) and explore the [prompt library](https://doutorai.pt/en/prompts).

## Frequently asked questions

### Is MCP a technology exclusive to Claude Code?

MCP was created by Anthropic but it is an open protocol. That means other AI assistants and tools can also adopt it. The specification is publicly available to encourage its adoption.

### Do I need to know how to code to use MCP?

To use existing MCP servers, you don't need to code. To create your own MCP server that connects a custom tool, you need development knowledge. But there are many ready-to-use servers for the most common tools.

### Is MCP safe? Does Claude Code have access to everything?

MCP works with a permissions model. Claude Code only has access to the tools and data you explicitly configure. There is no automatic access to anything: you decide what the assistant can use.

### What is the difference between MCP and a skill or agent?

Skills and agents define behaviors and tasks within Claude Code. MCP is about connectivity: it connects Claude Code to external tools and data. They are complementary concepts. An agent can use tools available via MCP.


---
Content by Doutor AI (https://doutorai.pt/en). When quoting, please cite the source: https://doutorai.pt/en/blog/what-is-mcp-in-claude-code