MCP (Model Context Protocol) is an open standard by Anthropic for connecting AI models to external tools and data. Think of it as USB-C for AI: one connection, all devices compatible.

How is MCP different from function calling?

Function calling is an API feature of a specific LLM provider. MCP is an open protocol that works with any client (Claude Desktop, Cursor, etc.) and any tool server. Function calling locks you to a vendor; MCP makes tools portable.

How is MCP different from LangChain?

LangChain is a framework you embed in your Python/JS code. MCP is a runtime protocol: a client app talks to a tool server over a standard wire format, no framework integration needed. MCP is simpler, language-agnostic, friendly to end-user clients like Claude Desktop.

How do I start using MCP?

The simplest way: open unyly.org, pick an MCP from the catalog, hit install. No JSON editing, no terminal. Works with Claude Desktop, Claude Code, Cursor and VS Code.

▸ explainer

What is MCP?

Model Context Protocol — Anthropic's open standard for AI agents. Explained simply.

▸ TL;DR

MCP = USB-C for AI. One protocol, any client + any tool server. The simplest way to start: open Unyly and install an MCP in one click.

Open Unyly catalog

The simple explanation

MCP (Model Context Protocol) is an open standard that lets AI models (like Claude) talk to external tools and data through one unified protocol. Anthropic released it in late 2024 and it has quickly become the de facto standard.

Analogy: think USB-C for AI. Used to be every device had its own cable; now one port for everything. Same with MCP — one connector between AI model (Claude, Cursor, VS Code) and any external tool (GitHub, Notion, Slack, databases).

How it works

MCP client — the app you chat with AI in (Claude Desktop, Cursor, Claude Code, VS Code).
MCP server — a separate program exposing tools (e.g. "send Slack message", "read Notion page").
Protocol defines how client and server talk: JSON-RPC over stdio or HTTP/SSE.

The AI agent in the client sees a list of tools from the MCP server and can call them. Server does the work, returns the result. That\'s it.

MCP vs function calling

Function calling is a specific LLM provider\'s API (OpenAI, Anthropic). You write code describing tools in their format and wire them to the LLM directly.

MCP is an open protocol. You write the server once and it works with any MCP client. No vendor lock-in.

MCP vs LangChain / LlamaIndex

LangChain is a framework you embed in your Python/JS code. Good for custom agents, but requires development work.

MCP is a runtime protocol. End-user clients (Claude Desktop) talk to tool servers directly, no framework integration needed. Simpler, language-agnostic, better for consumer AI apps.

How to start

The simplest path:

Open the Unyly catalog — 15,000+ MCPs.
Find the one you need (GitHub, Notion, Slack, Postgres, etc.).
Hit Install — pick your client tab (Claude Desktop / Code / Cursor / VS Code).
Restart the client. MCP connected.

Alternatively, you can hand-edit claude_desktop_config.json, but that\'s for enthusiasts only. Full install guide.