What does MCP stand for?

Anthropic introduced MCP and open-sourced it in November 2024 (the MCP guide covers its history and security), and the project’s own analogy is a USB-C port for AI applications: one standard plug instead of a custom connector per app.

Each word names part of the job. Model is the language model inside an AI app. Context is what that model can be given, which MCP splits into three things a server can offer: tools (functions the model can call), resources (data such as file contents or database records) and prompts (reusable templates). Protocol is the agreed message format, JSON-RPC 2.0, so any client can talk to any server.

The specification names three roles. The host is the AI application. It creates one client for each server it connects to, and each server is a program that provides context. Local servers run as a subprocess over the stdio transport (standard input and output); remote servers use Streamable HTTP (HTTP POST, with optional Server-Sent Events for streaming). The spec is versioned by date, and the current revision is 2026-07-28.

The specification’s own example: a client has listed a weather server’s tools with tools/list and found weather_current. That one definition is 89 tokens on o200k_base, and the host passes it to the model as an ordinary function calling tool.

When the model asks for the weather, the client sends the message below. Since revision 2026-07-28 every request carries its protocol version and the client’s capabilities in _meta, so the server needs no earlier handshake. The server answers with a content array, and the host hands that back to the model as the tool result.

{
  "jsonrpc": "2.0",
  "id": 3,
  "method": "tools/call",
  "params": {
    "name": "weather_current",
    "arguments": { "location": "San Francisco", "units": "imperial" },
    "_meta": {
      "io.modelcontextprotocol/protocolVersion": "2026-07-28",
      "io.modelcontextprotocol/clientInfo": { "name": "example-client", "version": "1.0.0" },
      "io.modelcontextprotocol/clientCapabilities": { "elicitation": {} }
    }
  }
}

A server written once works in every host that speaks MCP, instead of one integration per app. The cost is context: each connected server’s tool definitions take tokens. Claude Code defers MCP tool definitions by default and loads them only when needed.

Tools are code that can act on your behalf. The spec says hosts must get the user’s consent before invoking any tool, and that tool descriptions from untrusted servers should be treated as untrusted. Install servers the way you would install packages.

MCP vs function calling

Function calling is the model’s side: asking for a tool in one API. MCP is the app’s side: finding and running tools that live in separate servers. The two work together.

MCP vs an API

An MCP server often wraps an existing API, such as GitHub’s, and describes it in a form any AI app can discover and call. The API still does the work.

• Free toolMCP Config Generator→Build MCP server configs for Claude, Cursor, VS Code and Devin Desktop (Windsurf).
• Free toolAI Token Counter→Count tokens for GPT, Claude, Gemini, DeepSeek, Qwen and more.
• Guide · 11 min readWhat is MCP (Model Context Protocol)? A plain-English guideWhat the Model Context Protocol is, how hosts, clients and servers talk, what tools, resources and prompts do, which apps support it and how to use it safely.
• Guide · 12 min readHow to set up MCP servers in Claude CodeAdd MCP servers to Claude Code with claude mcp add, share them in .mcp.json, keep tokens out of git, pick the right scope and fix servers that won’t connect.

• Function callingFunction calling, also called tool use, is an LLM API feature in which the model replies with a structured request to run a function you described, with JSON arguments, which your code executes before sending the result back.
• AI agentAn AI agent is a program in which a language model works towards a goal by repeatedly choosing a tool to call, reading the result and deciding the next step, until the task is done or a stop condition is reached.
• SubagentA subagent is a separate AI worker that a main agent hands a task to, running in its own context window with its own system prompt, tools and model, and returning only its result to the main conversation.
• TokenA token in AI is the unit of text a language model reads and writes, usually a whole word, part of a word or a punctuation mark, which the model sees only as a number from its vocabulary.

All 40 terms in the AI glossary