AIDC-AI.IO MCP server

This is a helper that connects your AI assistant to the AIDC-AI.IO Design Engine, a service that plans AI data centers. You tell your AI things like how much power you need and how dense your racks are, and it works out the numbers for you. It is handy for people who plan or review data center projects in Korea and want a fast first pass without doing all the math by hand.

What this MCP server does

You type a request in plain words, like how many megawatts you need and how much power each rack uses. Your AI sends that to this helper, and the helper talks to the AIDC-AI.IO engine. The engine runs its own calculations and sends back rack counts, power figures, cooling split, cost, and a timeline. Your AI then shows you those results in the chat, along with any warnings. You can also ask it to check a design for problems or to draw a rack and site layout.

Click to zoom

What you can do with it

• Size an AI data center from an IT load, rack density, and GPU generation
• Get a design PUE, total MVA at 22.9 kV intake, and liquid versus air cooling split
• See CDU planning values, cost in Korean won, and a build timeline
• Validate a design against electrical, cooling, layout, safety, and data rules
• Get severity-classified findings and open questions (RFIs) for a design
• Generate a rack-plan grid with hall dimensions and rack positions in millimeters
• Generate a site-block layout for a chosen location

Try asking your AI

• “Size a 30 MW AI data center with 120 kW racks using Rubin GPUs on a 5000 square meter site in the Seoul area, with N+1 redundancy and liquid cooling, PUE target 1.2”
• “Check this design for problems: 30 MW IT load, 120 kW per rack, Rubin generation, 5000 square meters, metropolitan region”
• “Draw a rack plan and site layout for that same design, centered near Seoul”
• “What is the total MVA and liquid cooling share for a 50 MW Blackwell build with 80 kW racks?”

What it gives back to you

You get back a short summary with the main numbers, like rack count, design PUE, total MVA, cooling split, CDU count, cost in Korean won, and a timeline. For validation, you get a list of findings with severity levels and any open questions. For layout, you get a rack plan with hall sizes and rack positions in millimeters, plus a site-block layout. There may also be warnings and an engine version so you know which result you are looking at.

Before you start

Who it's for

Data center planners, engineers, and consultants who work on AI data center projects in Korea and want quick sizing, checks, and layouts.