Model Context Protocol
What is MCP (Model Context Protocol) and what an MCP server does
MCP (Model Context Protocol) is an open-source standard that connects AI applications to external systems such as files, databases, tools and services. An MCP server exposes tools, resources and prompts, and the application, for example Claude Code, opens one client for each server it connects to.
Updated
What is an MCP server?
A program that provides context to an AI application. It can expose:
- tools: functions the application can call, such as a database query;
- resources: context data, such as the database schema;
- prompts: reusable templates for working with the model.
How does MCP work?
- The host is the AI application, for example Claude Code, Claude Desktop or a code editor.
- The host creates one MCP client per server, each with a dedicated connection.
- Messages between client and server use JSON-RPC 2.0.
- A local server runs on the same machine and talks over stdio. A remote one connects over Streamable HTTP.
MCP vs API: what is the difference?
MCP standardizes how an AI application connects to external systems, and its docs compare it to a standard port for AI applications. The client asks the server for its list of tools and then calls them, so the application finds out at runtime what it can do. An MCP tool can in turn make API calls.
MCP or skill?
An MCP server gives access to a system, with its data and actions. A skill gives instructions on how to do a job. When a command-line tool exists for the same system, the Claude Code docs note that it uses less context than an MCP server.
What are the risks of a third-party MCP server?
- The Claude Code docs ask you to verify that you trust each server before connecting it.
- Servers that fetch external content can expose you to prompt injection.
- In interactive sessions, servers from a project's .mcp.json ask for approval before they are used.
- Servers you don't use can be disabled from /mcp.
Frequently asked questions
How do I add an MCP server to Claude Code?
With claude mcp add, using the http transport for a remote server or stdio for a local one. The default scope is local. Project scope writes the configuration to .mcp.json, which the team gets through version control.
Local or remote MCP server?
Local when the server runs on your machine and talks over stdio, usually to a single client. Remote when it runs at a provider and talks over Streamable HTTP, usually to many clients.
Which applications support MCP?
Assistants such as Claude and ChatGPT and editors such as Visual Studio Code and Cursor, among others.
How much context do MCP servers take up?
In Claude Code, MCP tool definitions are deferred by default, so only their names enter context until Claude uses a tool. The /context command shows what is taking up space.
Sources
Further reading
Related articles
Claude Code skills: instructions that load when needed
A Claude Code skill is a folder with a SKILL.md file that holds a name, a description and instructions. The description is always in context, so Claude knows which skills it has. The rest loads only when you run it with /name or when your request matches the description.ReadWhat is RAG in AI, and how does it work step by step?
RAG (Retrieval-Augmented Generation) means the model searches your documents first and writes the answer from what it found. The documents are split into chunks, each chunk becomes an embedding in an index, and for every question the chunks closest in meaning are added to the prompt next to the question.Read
Case studies
Related services
Want agents connected to your company's systems?
Tell us which systems your team works in. We will work out together what to connect over MCP and what access the agent gets.
Write to us