AI and process

MCP (Model Context Protocol)

Open protocol for AI tools

MCP (Model Context Protocol) is an open standard that lets AI applications and agents connect to external tools, data sources and business systems.

Official site

What is MCP (Model Context Protocol)?

MCP (Model Context Protocol) is an open protocol originated by Anthropic that defines how applications using language models connect to external data sources and tools. An MCP server exposes the resources and functions of a given system, such as a database, a shop API or documentation, and an AI application, such as an assistant or agent, connects to it as a client and uses them in a standardised way. Instead of building a separate integration for every AI application, a system is exposed once, through a single protocol.

How we use it at Koda Plus

We use MCP together with AI agents when analysing a company’s business logic: the protocol gives agents standardised access to the tools and data a given task requires. In implementations we can expose data and operations from a shop backend, for example one built on Medusa.js, through an MCP server, so that an AI assistant works within a strictly defined set of permissions.

When it makes sense

  • An AI agent needs data from several business systems
  • One tool should be available to different AI assistants
  • You need control over what a model can access
  • You are building internal tools powered by AI

Frequently asked questions
MCP (Model Context Protocol)

  • What is an MCP server used for?

    An MCP server gives an AI model access to specific resources and functions of one system, for example reading orders, searching documentation or writing to a database. An assistant or agent connects to it as a client and decides when to use each tool, within the permissions it has been granted.

  • Does MCP only work with Claude?

    No. Anthropic originated the protocol, but it is open, and its further development has been handed to a foundation operating under the Linux Foundation. It is supported by many AI applications and developer tools, including those from other providers, so an MCP server built once can work with different clients.

  • Is MCP secure?

    The protocol does not guarantee security by itself – that depends on the permissions given to the server and on who can use it. Good practice means minimal access, proper authentication, approval for operations that write data and using trusted servers only, because content they fetch may contain instructions designed to manipulate the model.

Free 15-minute call

Tell us about your project,
we'll tailor the rest to you

A 15-minute call is enough for us to know whether we are a fit for your project.
We start from what you need, not from an off-the-shelf solution.

Contact us

Remigiusz in a navy shirt and glasses
I use iMessage I use WhatsApp
Adrian in a grey turtleneck and glasses
I use iMessage I use WhatsApp