Source: https://arara.code.app.br/en/articles/what-is-mcp
Language: en-US
Published: 2026-10-31T20:25:00-03:00

Guides and comparisons

# MCP (Model Context Protocol): what it is and how to use it

MCP (Model Context Protocol) is an open standard for connecting AI applications to external systems, such as files, databases and search tools. Anthropic released it as open source on November 25, 2024. To use it, you register an MCP server in your AI assistant, which can then use the tools, resources and prompts that server offers. Stick to servers you know and give them minimal access.

Arara Code

Published and sources checked on October 31, 2026.

Published by the Arara Code project based on the official documentation cited.

## What MCP is and who created it

Anthropic announced the Model Context Protocol (MCP) on November 25, 2024, as an open-source project from day one. According to the announcement, David Soria Parra and Justin Spahr-Summers created the protocol at Anthropic. The announcement starts from a problem: every new data source needed its own custom integration, which made it hard to connect AI assistants to the systems where data lives. MCP replaces those one-off integrations with a single protocol.

The official MCP documentation compares the protocol to a USB-C port for AI applications: a standard way to connect AI to external systems. With it, assistants such as Claude and ChatGPT can reach data sources like local files and databases, tools like search engines, and workflows like specialized prompts. Development tools such as Cursor support it too. For developers, the benefit is building an integration once and using it in many clients.

[Anthropic: Introducing the MCP (Nov 2024)](https://www.anthropic.com/news/model-context-protocol)

## How MCP works: hosts, clients, servers and transports

MCP follows a client-server architecture. The host is the AI application, such as Claude Code or Claude Desktop. For each MCP server, the host creates one client, which keeps a dedicated connection to that server. The server is the program that provides context, and it can expose three core primitives: tools, functions the AI can run; resources, data that adds context; and prompts, reusable templates for interacting with the model.

Messages use JSON-RPC 2.0 and travel over one of two transports. With stdio, the server runs on the same machine as a local process and talks through standard input and output. Streamable HTTP lets you reach remote servers: the client sends messages with HTTP POST. This transport supports standard HTTP authentication, such as bearer tokens, API keys and custom headers, and the documentation recommends OAuth to obtain tokens. A local server usually serves a single client; a remote one usually serves many.

[MCP docs: Architecture overview](https://modelcontextprotocol.io/docs/learn/architecture)

Related

[Claude Code, Codex and Kiro in one editor: how to use themSee how to use Claude Code, Codex, Kiro and OpenCode in one IDE: official setup, each provider's account, approvals in the side chat and a comparison test.](https://arara.code.app.br/en/articles/claude-code-codex-kiro-in-one-editor)

## MCP security: what the specification recommends

A local MCP server is a program that runs on your computer with the same privileges as the client. The security best practices page in the specification lists the risks of servers from untrusted sources: running any command, data loss and data exfiltration. That is why a client that offers one-click server setup must show the exact command before running it and ask for your approval. Install only servers you know, and read that command.

The same page asks for least privilege. Clients should launch servers with restricted access to files, the network and other system resources, and tokens should carry minimal scopes, with no wildcard or full-access scopes. A server must never accept tokens that were issued for another service. Prompt injection is a further risk: when a tool reads third-party content, such as a web page, that text can carry hidden instructions for the AI. Review every action before you approve it.

[MCP specification: Security Best Practices](https://modelcontextprotocol.io/specification/2026-07-28/basic/security_best_practices)

## How to use MCP in Arara Code, step by step

In Arara Code, you register MCP servers only once. Open Arara settings › MCPs and write the configuration in JSON, with the “mcpServers” key. A local server takes command, args and env; a remote server takes url and headers. When you save, Arara Code validates the JSON and syncs the servers to Codex, Claude Code, Kiro and Google Antigravity. Each external assistant needs its own CLI installed and its own sign-in.

After saving, click “Check connection”: Arara Code starts the server and lists its tools. For a first test, try the filesystem server, which the official MCP documentation uses as an example of a local server running over stdio. Give it a test folder, never your home folder. Pass tokens only through env or headers, never in the chat, with the smallest scope that works. The chat shows the tool calls and asks permission before editing files or running commands.

[Arara Code: public IDE repository](https://github.com/arara-ide/arara-code)

## Sources and further reading

- [MCP docs: What is the Model Context Protocol?](https://modelcontextprotocol.io/docs/getting-started/intro)
- [Anthropic: Introducing the MCP (Nov 2024)](https://www.anthropic.com/news/model-context-protocol)
- [MCP docs: Architecture overview](https://modelcontextprotocol.io/docs/learn/architecture)
- [MCP specification: Security Best Practices](https://modelcontextprotocol.io/specification/2026-07-28/basic/security_best_practices)
- [Arara Code: public IDE repository](https://github.com/arara-ide/arara-code)

[Read as Markdown](https://arara.code.app.br/en/articles/what-is-mcp.md)

## Keep exploring

### [AI-generated code security: data and a checklist](https://arara.code.app.br/en/articles/ai-generated-code-security)

AI-generated code security: what Veracode, GitGuardian and research on invented packages show, plus a checklist based on the OWASP Top 10.

[Read the guide](https://arara.code.app.br/en/articles/ai-generated-code-security)

### [Coding with AI for beginners: a step-by-step guide](https://arara.code.app.br/en/articles/coding-with-ai-for-beginners)

Coding with AI for beginners: what 2025 data says to expect, a first project, basic Git, a safe routine and your first steps in Arara Code.

[Read the guide](https://arara.code.app.br/en/articles/coding-with-ai-for-beginners)

## Make room for your next idea.

Get to know Arara Code in your own project. Start with the installer and review the changes your assistant makes.

Download Arara Code
