How AI and MCPs work

When you chat with Claude, ChatGPT, or Gemini you are talking to a Large Language Model (LLM) — a very large neural network that has learned the statistics of human writing. By itself, an LLM is a frozen text predictor with a knowledge cut-off date. It cannot tell you what train is leaving Connolly in 4 minutes, or whether it will rain in Galway tonight.

The Model Context Protocol (MCP) — an open standard published by Anthropic — solves this. It gives AI assistants a common, secure way to call external tools in real time, much like USB standardised how devices plug into computers. IrishMCP hosts MCP servers for Ireland's most useful open datasets so your AI assistant can answer live questions grounded in Irish data.

This guide covers, in plain English:

• How LLMs work and why context is the key lever
• What MCP is and how a tool call flows end to end
• Where IrishMCP fits in — trains, weather, property, energy, Oireachtas and more
• Three worked examples (trains, weather, property)
• A full glossary and next steps