SSV.ASIA Start monthly

Learn AI · MCP

MCP, the AI plug-in standard.

You keep seeing "MCP". Here is what it is, why every modern AI tool is adopting it, and how it powers cited answers and real actions.

What MCP is in one sentence

MCP — Model Context Protocol — is an open standard that lets an AI model connect to external tools and data. Think of it as the USB-C of AI: one standard plug that connects any model to any tool, instead of a thousand custom wires.

Before MCP, every integration was custom. With MCP, a model can ask a tool "search the code vault", "read this file", or "run this check" through a common interface.

Why it matters

  • Models stopped guessing. An MCP can hand the model the actual building-code source, so the answer is cited, not inferred.
  • One tool works everywhere. The same MCP server works inside Claude, ChatGPT, or any MCP-aware client.
  • You own your tools. Open standard, no vendor lock-in.

A real example: the AEC code vault

The ssv.asia AEC MCP plugs a multi-million-word architecture/engineering corpus into Claude or ChatGPT. Ask "what is the seismic R-factor for an office under ASCE 7-22?" and the MCP retrieves the relevant source and returns a cited answer — with the 1 query/day free, more at ₹1 a query.

That is MCP in action: the model stays the brain, the MCP supplies the memory and the receipts.

Building codes, cited, inside Claude — in three steps

The ssv.asia MCP page says it best: "Technical building queries, answered with grounding. Inside Claude and ChatGPT via connectors." No new app to learn:

  1. Sign in to ssv.asia.
  2. Connect the MCP connector to Claude or ChatGPT (one click).
  3. Ask a building question — the answer comes back grounded in the knowledge base.

What's inside the knowledge base: the 12M-word AEC corpus — building codes, standards, material data and technical references, chunked and indexed for retrieval. Your AI does the thinking; the vault supplies the receipts.

MCP and the terminal

MCP also powers what a terminal can do — search your Obsidian vault (/lib), fetch the web, reach into models, run experiments. The more tools you attach, the more an agent becomes a real assistant rather than a chat window.

New to the ideas here? The glossary defines everything. Want the hands-on start? Your first hour with an AI terminal.