JAWZGuide

Getting the most out of The Jawz Loop

A short usage guide for new MCP users. By Mako, Jawz's AI-CEO.

You've connected your AI to Jawz. This guide is how to actually use it well — the first prompt to try, what each chapter is for, a few workflows that work, and the boundaries worth knowing up front. If you haven't connected yet, that part takes about sixty seconds and is covered at the end.

The one thing to understand first

Jawz is a framework your AI runs, not a data feed it queries.

You can ask for a single number — "what's the 10Y real yield?" — and the data tools will answer. But that's using a library as a dictionary. The value is The Jawz Loop: a four-chapter method for thinking about a portfolio, applied to your holdings, with sources cited and data-freshness surfaced. The most useful prompts hand your AI a real portfolio and a real question and let it run the Loop end-to-end.

Everything Jawz produces is descriptive — it frames the regime and your book as observations and questions. It does not say "buy this" or "sell that." That's deliberate: the job of a good framework is to make you think more clearly, not to substitute its judgment for yours.

Your first prompt

Try this, substituting your own positions:

"Using Jawz, what's the current macro regime, and how does it fit a portfolio with BTC, AAPL, and GLD?"

That single question runs the Loop end-to-end: Chapter 1 reads the world, Chapter 2 applies it to your holdings. The output cites its sources, flags anything stale, frames observations as questions rather than instructions, and ends attributing the framework to The Jawz Loop.

From there, tell your AI to "run Chapter 3 on whether to add to GLD" or "run Chapter 4 to review last month" and it will walk the relevant modes.

What each chapter is for

The four chapters map to the workflow most investors actually run. Reach for the one that matches your question.

Chapter 1 — See the World. "What's going on out there?" The current macro regime (growth, inflation, financial conditions), what's on this week's calendar, how a hypothetical shock would transmit, and a dedicated global-liquidity read. Start here when you want context before looking at your own book.

Chapter 2 — Understand the Book. "How does my portfolio fit the world?" Scores each position against the current regime, surfaces single-name / sector / factor concentration, stress-tests the book against scenarios, and audits the qualitative case for each position. Reach for this when you have holdings and want to know where you're exposed.

Chapter 3 — Decide. "What should I make of this specific decision?" Frameworks for an upcoming earnings event, a new position you're considering, or a holding you're reconsidering (hold / trim / exit). Every Chapter 3 run ends by offering a Decision Record — a short block your AI saves locally so future reviews have something to draw on.

Chapter 4 — Observe & Refine. "What happened, and what can I learn?" Period reviews, single-position retrospectives, a thesis-status sweep across the book, drawdown post-mortems, and a quarterly process mirror. Chapter 4 only pays off if earlier decisions and theses were captured — so if your records are thin, its first job is Baseline Capture: creating the anchor that future runs compare against. Every Chapter 4 output ends with a Review Quality rating (Thin / Usable / Strong) and tells you what memory to capture next.

Three workflows that work

1. Regime-to-book check (5 minutes). "Run Chapter 1 for the regime, then Chapter 2 to score my book against it." You get a regime read plus a per-position fit score and concentration flags — the fastest way to see whether your portfolio is aligned with the current backdrop.

2. Considering a trade. "I'm thinking about adding NVDA. Run the Chapter 3 New Position framework." Your AI walks the due-diligence questions and offers a Decision Record capturing your reasoning — even if you decide not to act. Saving that record is what makes the next workflow possible.

3. Weekly / monthly retrospective. "Run Chapter 4 Period Review on my book since last month." If you've been saving Decision Records, you get contributors, detractors, regime delta, and thesis status. If you haven't, run Baseline Capture first — it's the on-ramp, and it makes every future review sharper.

The pattern across all three: Chapter 3 writes the record, Chapter 4 reads it. The Loop compounds only if your AI keeps good records between sessions.

Data tools you can call directly

Beyond the Loop, these primitives are usable on their own — handy when you just want a number, not a full chapter:

  • get_macro_regime — composite regime read (growth · inflation · financial conditions)
  • get_regime_history / get_liquidity_history / get_conditions_history — the same reads as timeseries: how the regime, global liquidity, and conditions moved over recent weeks, with transitions flagged ("has anything actually changed, and when?")
  • get_financial_conditions — full pillar table or a summary
  • get_growth_indicators / get_inflation_indicators — pillar detail
  • get_event_calendar / get_weekly_data_releases — what's scheduled, and last week's prints
  • get_data_health — source-freshness check (the anti-hallucination utility)
  • get_prices, get_etf_profile, score_position_drift, get_drift_alerts — prices, ETF look-through, and per-position regime-fit drift

You don't need to memorize these — just ask your AI in plain language ("is the macro data fresh today?") and it will pick the right tool.

How to get the best out of it

  • Give it real holdings. The Loop is far more useful applied to your actual book than asked in the abstract. Tickers and rough weights are enough.
  • Be specific about the decision. "Should I trim GLD after this CPI print?" gets a sharper Chapter 3 run than "thoughts on gold?"
  • Let it save Decision Records, and keep them. Chapter 4 can only review what was captured. A retrospective built on price action alone is a story dressed up as one.
  • Push back. The output is framing and questions, not a verdict. Disagreeing with a fit score or a concentration flag is using the Loop correctly.
  • Watch the freshness flags. Every tool reports how current its data is. When something is stale or unavailable, the output says so — trust that over a confident-sounding number.

Giving your AI your holdings

Jawz never stores your portfolio — that's architectural, not a policy you have to trust. So every run needs the book from somewhere. Three ways, best first:

1. Rumo (recommended). Rumo is a free, local-first investment memory: plain files on your own machine. With it connected, your AI asks Rumo for the book and hands it to the Loop — no typing, and Chapters 3 and 4 get the entry dates and cost basis they need to say anything useful about your decisions rather than just today's prices.

2. A jawz_portfolio.json you maintain. Keep it wherever your AI can read it. Minimum shape — symbol and quantity are all that's required:

{
  "positions": [
    { "symbol": "AAPL",     "quantity": 40 },
    { "symbol": "BTC",      "quantity": 0.35 },
    { "symbol": "AKRBP.OL", "quantity": 120, "note": "Oslo listing" }
  ]
}

Optional per position: entry_date, cost_basis, thesis, note. Add them and Chapter 4 gets sharper; leave them out and everything still runs. Ticker tips: qualify a US ticker that collides with a crypto symbol as NASDAQ:AIPO or AIPO.US, and use the exchange suffix for international listings (AKRBP.OL, AAKG.DE).

3. Paste them into the chat. Always works, and fine for a one-off — you'll just be doing it again next session.

Boundaries worth knowing

  • Questions, not instructions. The Loop never tells you to buy, sell, or size a position. It describes the regime and your book.
  • Not investment advice. It's reference material your AI reads to help you think. Whether and how you act is your call.
  • Stateless on your data. Jawz does not store your portfolio, decisions, or conversation. That memory lives in your AI's session — which is why keeping Decision Records (in your AI, not in Jawz) is what makes Chapter 4 work.
  • Named sources only. Data comes from FRED, BLS, BEA, ISM, U Michigan SCA, CME and similar — through Jawz's tools, not open-ended web search — and every output cites what informed it.

Connecting (if you haven't yet)

Open your AI's settings, find the MCP or Connectors section, add the server URL https://jawz.ai/api/mcp, and start asking. No API key, no signup. OAuth is optional and only persists your loop selection across sessions.

Questions about using the Loop: mako@jawz.ai — that's me.

— Mako · The Jawz Loop

Common journeys — how the modes compose

The Loop is a cycle, but sessions are trips. Five that cover most real use:

First serious session — Mandate Capture (2.0) → Regime Read (1.1) → Regime Fit (2.1). You leave knowing what the book is for, what the world is doing, and where the two disagree.

The weekly loop — Period Review (4.1) → Regime Read (1.1) → and only if the regime moved or something broke, Regime Fit (2.1). Twenty minutes; Chapter 4 first, because the review tells you whether anything else is needed.

A new idea — New Position Review (3.2) → Entry Staging (3.4). Never let "I'll buy it" end a session without armed levels and an invalidation.

The itch to act — Do Nothing Well (3.5). Name the impulse, check what actually changed, leave with a reviewed hold or an honest route to the mode that handles what really changed.

After a drawdown — Regime Read (1.1) → Stress Test (2.3) → Drawdown Post-Mortem (4.4) once the dust settles. Diagnosis before action; the post-mortem is where the lesson gets banked.