@okaneland/mcp

A small Model Context Protocol server that gives AI agents two things from Okane Land:

It runs on the user's machine over stdio, so there is nothing to host. get_ai_tool_prices fetches the public /api/v1/prices.json at call time, so the data is always as fresh as the site's monthly re-verification.

Install

Add it to your MCP client's config (Claude Desktop, Cursor, Windsurf, and others use the same shape):

{
  "mcpServers": {
    "okaneland": {
      "command": "npx",
      "args": ["-y", "@okaneland/mcp"]
    }
  }
}

Restart the client. The two tools appear in any agent you run there.

Example

You: I want a $20/mo AI writing tool on Claude Sonnet. What do similar tools charge, and would $20 keep money at 500 users?

The agent calls get_ai_tool_prices and calculate_ai_margin, then answers from real data: comparable tools charge more, and at 500 users on a Sonnet-class blend you keep about $3,260/mo (33%), staying profitable down to a $12.25 price or up to 3.3M tokens per user.

Tool reference

get_ai_tool_prices

Inputs (all optional): category, query, id, limit. Returns the matching tool rows plus the index's updated date, method, and source page.

calculate_ai_margin

Inputs: pricePerMonth, payingCustomers, blendedTokenPricePerM, tokensPerUserM, and optional monthlyChurnPct (6.1), intlCardPct (40), vatPct (0), failedPaymentsPct (9), refundsPct (3). Returns gross, tokenBill, cardFees, failedPayments, refunds, tax, keep, keepPct, perCustomerKeep, customersReplacedMonthly, breakEvenPrice, breakEvenTokensPerUserM, and tokenCostExceedsPrice.

Privacy

The server fetches only the public price index over HTTPS and sends nothing about you. The margin calculation runs entirely locally.

Develop

npm install
npm run build     # tsc -> dist/
npm test          # unit tests pin the margin math to the web calculator
npm run smoke     # spins up the server, runs the MCP handshake, calls both tools

The margin math in src/margin.ts is a port of Okane Land's margin calculator; src/margin.test.ts pins its known-good outputs so the two cannot silently drift.

Disclaimer

Numbers are directional, not financial advice.

License

MIT