MCP integration

Plug Kubova into Claude, Cursor, Cline, ChatGPT, or any MCP-aware AI agent. Pack containers with a single tool call.

1Connect by URL (no install)

Add Kubova as a remote connector with one URL. Your AI client opens a secure sign-in to your Kubova account — no API key, no terminal.

https://kubova.com/mcp

Claude (claude.ai / Desktop): Settings → Connectors → Add custom connector → paste the URL.
ChatGPT: Settings → Apps & Connectors → enable Developer mode → Create → paste the URL.
Cursor · VS Code: Add the URL as a remote MCP server; sign in when prompted.

Tool calls run on your plan and quotas. Pro or Team required for packing and reports (14-day trial available).

2Talk to it

Restart your client, then ask the AI anything that needs container packing:

> Pack 20 boxes of 50x40x30 cm into a 40HC container.
> What's the utilization?

The agent will call the pack_containers tool, render placements, and explain the result.

Alternative: local install (npm)

Prefer a local stdio server (offline config files, CI, n8n)? Use the npm package with an API key.

Get an API key →

{
  "mcpServers": {
    "kubova": {
      "command": "npx",
      "args": ["-y", "@kubova/mcp@latest"],
      "env": {
        "KUBOVA_API_KEY": "kbv_..."
      }
    }
  }
}

Claude Code: ~/.claude/settings.json · Claude Desktop: claude_desktop_config.json · Cursor: ~/.cursor/mcp.json

Tools exposed

Tool	Purpose
pack_containers	Pack a list of SKUs into one or more containers. Returns per-container placements with x,y,z and utilization.
generate_report	Render a PDF report from a previous pack result. Returns a downloadable URL.
estimate_capacity	How many complete sets of a product (1+ carton types per set) fit in one container — real 3D packing, with the binding limit per container.
verify_key	Sanity check the API key — returns identity, scopes, and rate limit.

Source

The MCP server is open source on GitHub: Kubova-com/kubova-mcp.