SupliiChain MCP Sandbox

A lightweight, local inventory-intelligence MCP server. Run it on your machine in under 60 seconds and test how your LLM setup or code editor queries structured inventory schemas before you connect a production stack.

Read-only. Zero config. No account, no API keys, no database. The data is a deterministic sandbox catalog: the same SKU always returns the same numbers, which proves the server is computing answers rather than the model hallucinating them.

Quickstart

git clone https://github.com/AMinerSolution/supliichain-mcp-sandbox.git
cd supliichain-mcp-sandbox
npm install
npm start

The server speaks JSON-RPC 2.0 over Streamable HTTP at http://localhost:3333/mcp.

Connect Claude Code (or Cursor / Windsurf / VS Code)

claude mcp add --transport http supliichain-sandbox http://localhost:3333/mcp
claude mcp list   # should show "Connected"

Connect Claude Desktop

Add to claude_desktop_config.json:

{
  "mcpServers": {
    "supliichain-sandbox": {
      "url": "http://localhost:3333/mcp"
    }
  }
}

Try it

Ask your model:

What's my current stock on SKU-001, SKU-002, and SKU-003 across all warehouse nodes?

Which of my open purchase orders from VENDOR-01 are late against the promised lead time?

Pull 14, 30, and 90 day sales velocity for SKU-014 and tell me if demand is accelerating.

Tools

Tool	What it answers
get_stock_levels	On-hand, available, committed, and in-transit stock per warehouse node
get_velocity_metrics	Trailing sales velocity over any lookback windows
get_purchase_orders	PO status, expected arrival, and promised vs actual lead times

Sandbox catalog: SKU-001 through SKU-050, warehouse nodes US-WEST, US-EAST, US-CENTRAL, suppliers VENDOR-01 through VENDOR-04. Any SKU string works; the catalog ids are just a convention.

What this is a sandbox of

This is the free, local slice of the SupliiChain gateway. The hosted version adds 10 more tools behind the same protocol: ABC/XYZ classification, safety stock drift, dead stock identification, stockout horizon prediction, time-series forecasting (ARIMA / Prophet / AutoML ensemble), forecast error scoring, demand elasticity, marketing performance, and an executive P&L impact rollup.

Get a free hosted token (3 read-only tools, no credit card) or see the Operations Optimization Workshop at dpx-mcp.vercel.app.

Security posture

• Every tool input is a bounded, validated Zod schema (max lengths on all strings and arrays).
• Enum-constrained supplier and node ids reject anything outside the known set.
• Read-only by design: there are no write tools in this sandbox.
• Stateless transport: a fresh server instance per request, no sessions.

License

MIT