AIDC Design Engine Plugin

Size, validate, and lay out AI data centers from Claude or Codex - for NVIDIA Hopper, Blackwell (GB200 NVL72), and Rubin (Vera Rubin NVL72) deployments. Backed by the deterministic engineering engine at aidc-ai.io.

Install

Claude

claude plugin marketplace add aidc2026ai-melon/aidc-design-engine
claude plugin install aidc-design-engine@aidc-ai

Codex

codex plugin marketplace add aidc2026ai-melon/aidc-design-engine
codex plugin add aidc-design-engine@aidc-ai

What's inside

Component	Purpose
MCP server aidc-design-engine	Three tools: design (rack count, PUE, MVA, cost, schedule), validate (electrical/cooling/layout/safety rule QA with severity-classified findings and RFIs), and layout (rack plan in mm plus site plan blocks)
Skill aidc-design-engine	Encodes one canonical DesignRequest across design -> validate -> layout -> report, GPU-generation density/PUE defaults, validation gating, rack-count and MVA semantics, and a standard design-basis report template
Codex marketplace	Connects directly to https://aidc-ai.io/api/mcp; the Claude plugin continues to use npx aidc-mcp-server

Requirements

• Claude: Node.js >= 18 (the MCP server runs via npx -y aidc-mcp-server)
• Codex: no local Node.js requirement; it uses the Remote MCP endpoint
• Internet access to https://aidc-ai.io

Authentication (optional)

Anonymous use is allowed (10 requests/hour). The Codex Remote MCP transport maps the following environment variable to a Bearer token. Claude stdio also uses it when the installed aidc-mcp-server release supports bearer forwarding:

AIDC_API_KEY=aidc_live_...

Claude stdio key forwarding requires aidc-mcp-server 0.2.2 or later; npm latest is 0.2.2 as of 2026-07-14. The Codex Remote MCP transport forwards the same variable directly as a Bearer token.

Example prompts

Can I fit 60 MW of GB200 on a 20,000 m² site near Seoul?

Design a 50 MW Vera Rubin data center on a 15,000 m² regional site with liquid cooling, validate it, and give me the rack layout.

Compare Blackwell vs Rubin on the same 30,000 m² site at 2N redundancy.

Notes

• Engine results are a design basis, not a permit-ready design — local utility, AHJ, climate, water, seismic, and security assumptions remain project-specific validation inputs.
• Every response includes the engine version and a citation to aidc-ai.io.
• Terms: https://aidc-ai.io/terms · Privacy: https://aidc-ai.io/privacy

MIT © AIDC-AI.IO