MCP Explained

316 points · 133 comments on HN · read original →

Points and comments are a snapshot, not live.

MCP is an open standard for connecting AI apps to tools and data via JSON-RPC.

The Model Context Protocol (MCP) is an open standard where a server exposes tools, data, or prompts. The AI app (Claude, Cursor, ChatGPT) shows these to the model. The model requests actions, the app calls the server, and hands results back. Communication uses JSON-RPC over stdin/stdout or an HTTP endpoint, with defined methods, capability negotiation, transports, and auth.

What commenters are saying

Many commenters found the explainer too dense, convoluted, or lacking clear first-principles teaching. Some noted it assumes prior knowledge and is more reference than guide. Others defended its brevity. A recurring correction: MCP is JSON-RPC, not REST. Several suggested that in the age of LLMs, interactive learning via a chat model may be more effective than static explainers.