You:    Here's my IBKR trade history [trades.csv]. Set up firma.
Claude: I see 47 transactions. Logging in chronological order... done.
        Syncing prices and FX history... your portfolio: $179K, +61% all-time. TSLA is 78%.

You:    My portfolio feels really volatile lately. Should I be worried?
Claude: TSLA is 74% of your equity — a 10% TSLA move is a 7% net worth move.
        Cash flow is solid at $6.4k/month saved. The volatility is market, not lifestyle.

You:    That July expense spike looks alarming.
Claude: That was $57k in TSLL — a leveraged bet, not spending.
        Strip it out: expenses flat at ~$7k/month, savings rate 47.8%.

You:    Give me a full overview with charts.
Claude: Net worth: $294k → $490k in 8 months, pulled back to $427k with TSLA's Q1 dip.
        [builds net worth trend, income vs expenses, portfolio allocation, unrealized P&L]

Your brokerage app shows you a number. firma shows you the truth.

• It talks to Claude. Natively. Every command lives behind MCP — log trades, sync prices, render charts, reason about your money. You type plain English; Claude does the rest.
• Local-first, no exceptions. Everything sits in ~/.firma/firma.db. No server, no sync, no account. Your financial data never leaves your machine. Ever.
• Nine markets, one source of truth. US, KRX, KOSDAQ, JP, HK, LSE, XETRA, NSE, TW — plus crypto, commodities, and real estate. Stored in USD, shown in any currency, with historical FX from FRED: a 2018 balance gets the 2018 rate, not today's.
• No holdings table to lie to you. Portfolio, cost basis, P&L — all derived live from your trade log. Nothing drifts out of sync, because there's nothing to drift.
• A brief that actually has an opinion. get_market_brief doesn't dump data — it takes a stance. A weekly lean (aggressive or defensive), a ranked call per holding (BUY / ADD / HOLD / TRIM / SELL / AVOID with conviction and an English thesis), and a watchlist of what to look at next. Every holding is classified, run through a category playbook, and checked by a correlation guard that catches "more of what you already own" before it lands.
• A simulator nobody else ships. project_net_worth runs multi-scenario Monte Carlo over your portfolio at any horizon — next month or 30 years, one tool. simulate_scenario does the deterministic day-to-week version ("PCE comes in hot Thursday — how much does that cost me?"). You supply the assumptions, Claude discloses them verbatim, firma does the math. No hidden "advisor view."
• Time-series features from day one. The first firma sync backfills 90 days of daily history. Correlation, Sharpe, max drawdown, volatility, CAGR — populated immediately, not 30 days from now.
• It tracks your decisions, not just your positions. show_thesis_track reads your own trade reasons back to you against the outcome. show_anti_portfolio follows what you sold — the data brokerages hide because it exposes bad exits.

Get started

# 1. Install
npm install -g @evan-moon/firma

# 2. Set API keys (both free)
firma config set finnhub-key YOUR_KEY   # finnhub.io — prices, news, earnings
firma config set fred-key YOUR_KEY      # fred.stlouisfed.org — macro & FX history

# 3. Connect Claude Desktop, then restart it
firma mcp install

Rather not touch a terminal? Grab the macOS or Windows installer — Node.js is bundled, Claude is wired up for you.

Then drop your data into Claude — a CSV, a brokerage export, plain text. It logs everything and syncs prices automatically.

If firma helps you see where your money went, ⭐ star the repo — it's the cheapest way to help others find it.

How it's split: MCP thinks, the CLI keeps you safe

firma is MCP-first. Analysis, briefs, recommendations, news, valuation, projections — all of it happens in conversation with Claude. The CLI exists for the three things a chat can't safely own: setup, manual data entry, and repair.

# Setup
firma config set <key>        # finnhub-key / fred-key / currency
firma mcp install             # register with Claude Desktop
firma doctor                  # check keys, data state, FX cache, coverage

# Manual entry
firma add txn                 # buy / sell / deposit / dividend / tax — any of 9 markets
firma add monthly             # month-end balance + cash flow in one flow
firma add snapshot            # sync prices, snapshot today's portfolio

# Repair (when Claude or you got it wrong)
firma edit txn / balance / flow / snapshot
firma delete txn / ...        # alias: firma rm

# Verify (read-only — just confirm what landed in the DB)
firma show portfolio / txns / balance / flow / snapshot / profile / fx

# Sync
firma sync                    # prices (Finnhub) + FX history (FRED) + correlation cache
firma sync fx                 # FX history only

Everything analytical — get_market_brief, every show_*, report, project_net_worth, simulate_scenario — lives in Claude. Two tool families that never mix: diagnostic (what's true right now, no fortune-telling) and thought experiment (what-ifs, assumptions always disclosed).

Architecture

A Yarn Berry monorepo with a strict port-and-adapter layout. The application layer never imports an external API directly — it talks to domain interfaces, and adapters implement them.

packages/
  domain/        ports + types, zero external-API knowledge
  external-api/  raw clients (finnhub, fred, worldbank, gdacs, yahoo, …), zero domain knowledge
  adapter/       the only layer that imports both sides
  use-case/      business logic (portfolio, brief, advisor, projection, scenario)
  shared/        db + utils
apps/
  cli/  mcp/  docs/

The rule: external-API packages know nothing about firma's domain; domain packages know nothing about any external API. @firma/adapters is the single bridge. Adding a data source touches no command and no business logic — a new raw client, one adapter, one registration line.

See CONTRIBUTING.md for the full walkthrough and the recipe for adding a provider.

Development

Requires Node.js 22+ and Yarn Berry.

corepack enable
yarn install
yarn dev:cli show portfolio    # CLI dev mode
yarn typecheck
yarn test

Works with Herald

firma is the financial-intelligence layer of the Herald ambient voice stack. Connected, Herald reports your net worth by voice, answers "how much did I spend on fees this quarter?" without a spreadsheet, and pulls market context mid-conversation. Herald + firma + Memex: ambient voice, financial intelligence, persistent memory — one personal AI stack.

License

MIT © Evan Moon