💰 Wealth Engine MCP

An MCP server that helps you make more money and keep more of it — a complete personal-finance and business-economics toolkit your AI assistant can use to give you real, math-backed answers.

No API keys. No accounts. No network calls. Every answer is computed locally from battle-tested financial formulas, so it works the moment you install it and produces the same numbers every time.

The honest truth: software can't literally print money. But the decisions these tools inform — charging the right freelance rate, pricing a product at the profit-maximizing point, killing the avalanche of interest on your debt, picking the investment with the best risk-adjusted return — are worth a lot of money over a lifetime. That's where the leverage is.

Why this exists

Most "money" advice is vibes. This is arithmetic. Each tool implements the actual formula a financial analyst would use, returns a transparent breakdown (not just a single number), and is covered by tests that check it against known-correct values.

• ✅ 11 financial tools, each a real engine

• ✅ Zero setup — no keys, no DB, runs over stdio

• ✅ 58 passing tests validating the math against closed-form results

• ✅ Strict TypeScript, structured JSON output for both humans and agents

• ✅ Works in Cursor, Claude Desktop, and any MCP client

Related MCP server: plaid-mcp

The tools

Install

1. Build it

npm install
npm run build

This produces dist/index.js, the runnable MCP server.

2. Register it with your MCP client

Cursor

Create or edit .cursor/mcp.json in your project (or ~/.cursor/mcp.json for global use):

{
  "mcpServers": {
    "wealth-engine": {
      "command": "node",
      "args": ["C:/Users/jejej/OneDrive/Escritorio/github_projects/mcp/dist/index.js"]
    }
  }
}

Then open Cursor → Settings → MCP and confirm wealth-engine shows 11 tools. Ask the agent something like "Use the freelance rate tool to tell me what to charge for a $150k target."

Claude Desktop

Edit claude_desktop_config.json:

• macOS: ~/Library/Application Support/Claude/claude_desktop_config.json

• Windows: %APPDATA%\Claude\claude_desktop_config.json

{
  "mcpServers": {
    "wealth-engine": {
      "command": "node",
      "args": ["C:/Users/jejej/OneDrive/Escritorio/github_projects/mcp/dist/index.js"]
    }
  }
}

Restart Claude Desktop. The tools appear under the 🔌 menu.

Tip: During development you can skip the build step and run the TypeScript directly with "command": "npx", "args": ["tsx", "/abs/path/src/index.ts"].

Tool reference & examples

Every tool accepts a JSON object and returns both a human-readable text block and machine-readable structuredContent. Below are representative inputs and the headline outputs.

compound_growth

{
  "principal": 10000,
  "contribution": 500,
  "annualRatePct": 7,
  "years": 30,
  "periodsPerYear": 12,
  "annualInflationPct": 2.5
}

Returns futureValue, totalContributed, totalInterest, realFutureValue (today's purchasing power), growthMultiple, and a year-by-year schedule.

loan_amortization

{ "amount": 400000, "annualRatePct": 6.5, "years": 30, "extraPayment": 300 }

Returns the periodicPayment, totalInterest, payoffYears, interestSavedVsNoExtra, and a full amortization schedule.

debt_payoff

{
  "debts": [
    { "name": "Visa", "balance": 6000, "annualRatePct": 24.99, "minPayment": 150 },
    { "name": "Car",  "balance": 12000, "annualRatePct": 6.5,  "minPayment": 280 }
  ],
  "monthlyBudget": 900,
  "strategy": "compare"
}

compare (default) runs avalanche and snowball, reports monthsToDebtFree and totalInterestPaid for each, and tells you which is cheaper and by how much.

fire_calculator

{ "annualIncome": 120000, "annualExpenses": 48000, "currentInvestments": 80000, "realReturnPct": 5 }

Returns your fireNumber, savingsRate, yearsToFire, plus leanFireNumber and fatFireNumber.

freelance_rate

{
  "targetAnnualIncome": 150000,
  "annualBusinessExpenses": 12000,
  "taxRatePct": 30,
  "workingWeeksPerYear": 46,
  "billableUtilization": 0.6
}

Returns the hourlyRate and dayRate you must charge, the billableHoursPerYear, and how that compares to the naive salary / 2080 mistake (vsNaiveSalaryRate).

price_optimizer

{
  "demandPoints": [
    { "price": 29, "quantity": 420 },
    { "price": 39, "quantity": 300 },
    { "price": 49, "quantity": 190 }
  ],
  "unitCost": 8,
  "fixedCost": 2000
}

Fits a linear demand curve, then returns the optimalPrice (profit-max), revenueMaximizingPrice, the fitted demandModel with R², and a transparent price sweep.

roi_compare

{
  "investments": [
    { "name": "Rental",   "cashFlows": [-50000, 6000, 6000, 6000, 6000, 76000] },
    { "name": "Business", "cashFlows": [-50000, 0, 20000, 30000, 40000] }
  ],
  "discountRatePct": 8
}

Returns per-investment roi, paybackPeriod, npv, irrPct, and the winners bestByNpv / bestByRoi / bestByIrr.

tax_estimate

{ "income": 95000, "deduction": 14600 }

Defaults to US 2024 single-filer federal brackets (override with your own brackets for any country/year). Returns totalTax, afterTaxIncome, effectiveRatePct, marginalRatePct, and the per-bracket breakdown.

budget_analyzer

{
  "monthlyIncome": 6000,
  "categories": [
    { "name": "Rent", "amount": 2200, "type": "needs" },
    { "name": "Dining", "amount": 800, "type": "wants" },
    { "name": "Index funds", "amount": 1200, "type": "savings" }
  ]
}

Compares your split to 50/30/20, finds unallocated money and overspending gaps, and returns concrete recommendations.

subscription_audit

{
  "subscriptions": [
    { "name": "Streaming A", "amount": 15.99, "cycle": "monthly", "usesPerMonth": 12, "valueRating": 4 },
    { "name": "Gym",         "amount": 60,    "cycle": "monthly", "usesPerMonth": 1,  "valueRating": 2 }
  ]
}

Annualizes every line, computes costPerUse, flags each as keep / review / cut, and totals potentialAnnualSavings.

side_hustle_ranker

{
  "hustles": [
    { "name": "Consulting", "monthlyRevenue": 4000, "hoursPerMonth": 30, "successProbability": 0.9 },
    { "name": "Print shop", "monthlyRevenue": 2500, "monthlyCost": 600, "hoursPerMonth": 60, "startupCost": 5000, "rampMonths": 4, "successProbability": 0.6 }
  ]
}

Ranks by risk-adjusted effective hourly value, with breakEvenMonths and expectedYearOneProfit for each.

Development

npm run dev        # run the server from TypeScript (tsx)
npm run typecheck  # strict type checking, no emit
npm test           # run the full vitest suite
npm run test:watch # watch mode
npm run build      # compile to dist/

Project layout

src/
  index.ts            # stdio entry point (the MCP process)
  server.ts           # registers all 11 tools with Zod schemas
  finance/            # pure, dependency-free financial engines
    compound.ts  loan.ts  debt.ts  fire.ts  freelance.ts
    pricing.ts   roi.ts   tax.ts   budget.ts  subscriptions.ts
    sideHustle.ts  money.ts
test/                 # one spec per engine + an end-to-end server test

The design rule: all math lives in src/finance/ as pure functions with no MCP or I/O dependencies. The server is a thin adapter. That keeps the logic trivially testable and reusable.

How the numbers are validated

The tests don't just check "it runs" — they assert against independently-derived values:

• Compound growth is checked against the closed-form annuity future-value formula.

• Loan amortization is checked against the standard mortgage payment (a $100k/6%/30yr loan → ~$599.55/mo, ~$115,838 total interest).

• Tax is checked bracket-by-bracket against hand-computed US 2024 figures.

• IRR is checked against known multi-period rates and the trivial [-100, 110] → 10% case.

• Price optimization is checked against the analytic optimum of a known linear demand curve.

• An end-to-end test spins up the server over an in-memory transport, lists the tools, and calls them through the real MCP protocol.

Run npm test to see all 58 pass.

Disclaimer

This is an educational/decision-support tool, not financial, tax, investment, or legal advice. Tax brackets and assumptions are illustrative defaults — verify against current rules for your jurisdiction, and consult a qualified professional before making consequential decisions.

License

MIT — see LICENSE.