finance-calc-mcp (Node.js)

A local MCP server that provides basic arithmetic plus common finance calculations. Same calculations as the Python version, in Node. Uses decimal.js throughout — no floating-point errors in money math.

Install

cd finance-calc-mcp
npm install

That's it. No build step (this is plain ES modules, no TypeScript).

Verify it works:

npm test

You should see "All tests passed." Optionally, sanity-check the live MCP server against itself by running node probe.js, which spawns the server as a subprocess and exercises a couple of tools.

Wire into Claude Desktop

Edit claude_desktop_config.json:

• Windows: %APPDATA%\Claude\claude_desktop_config.json
• macOS: ~/Library/Application Support/Claude/claude_desktop_config.json

Add:

{
  "mcpServers": {
    "finance-calc": {
      "command": "node",
      "args": ["C:/absolute/path/to/finance-calc-mcp/src/index.js"]
    }
  }
}

Restart Claude Desktop. The 17 finance tools should show up.

Use from your own Node.js code

You don't need Claude in the loop — you can call the server directly from any Node program using the MCP client SDK. See probe.js for a working example. The pattern:

import { Client } from "@modelcontextprotocol/sdk/client/index.js";
import { StdioClientTransport } from "@modelcontextprotocol/sdk/client/stdio.js";

const transport = new StdioClientTransport({
  command: "node",
  args: ["./src/index.js"],
});
const client = new Client({ name: "my-app", version: "1.0.0" }, { capabilities: {} });
await client.connect(transport);

const result = await client.callTool({
  name: "loan_payment",
  arguments: { principal: 200000, annual_rate_pct: 6.0, term_months: 360 },
});
console.log(result.content[0].text);  // "1199.10"

await client.close();

Tools

Basic: add, subtract, multiply, divide, power

Loans:

• loan_payment(principal, annual_rate_pct, term_months)
• loan_amortization_summary(...)
• loan_remaining_balance(..., payments_made)

Interest:

• simple_interest(principal, annual_rate_pct, years)
• compound_interest(principal, annual_rate_pct, years, compounds_per_year=12)
• daily_interest_accrual(principal, annual_rate_pct, days, day_count_basis=360) — A/360 or A/365

Depreciation:

• depreciation_straight_line(cost, salvage_value, useful_life_years)
• depreciation_declining_balance(cost, salvage_value, useful_life_years, factor=2.0) — full year-by-year schedule

Tip: tip(bill, tip_pct=18, split_n=1)

Dividend MTM: dividend_mark_to_market(shares, dividend_per_share, cost_basis_per_share, current_price_per_share, dividends_received=0)

PV/FV:

• present_value(future_amount, annual_rate_pct, years, compounds_per_year=1)
• future_value(present_amount, annual_rate_pct, years, compounds_per_year=1)

Why Decimal not float

Money math in floats has well-known precision bugs (0.1 + 0.2 ≠ 0.3). Every calculation in this server runs through decimal.js with 28 digits of precision and explicit rounding at the boundary. Inputs come in as floats over the JSON wire — they're converted to Decimal on the way in so nothing accumulates float drift downstream.