AI Employees

Wealth-Ops

Private beta Does the work of a Wealth analyst

A full AI research desk for your portfolio that runs on your own hardware and tells you the truth, even when the truth is "no edge here."

Email us for a seat See Accounting-Ops
Research deskQuant engineGuarded execution

Who it is for

Built for advisors and principals who need the whole book in one place.

Self-directed investors and operators who want serious investment research without handing their portfolio to a cloud, and who would rather hear "no demonstrated edge" than a sales pitch.

Why it wins

Positions and plans, not dashboards for their own sake.

01

Your positions, research, and trade history stay on your hardware. No cloud LLM ever sees your portfolio.

Runs a full research desk on demand: five analysts, a bull/bear debate, a judge, a risk officer, and a portfolio manager, end to end on a local model.

02

Every agent call is logged at the price it was made and scored against what actually happened next. You get a measured hit-rate, not confident vibes.

Computes every number in a real quant engine (numpy, pandas, scipy). The agents reason over the math and cite it; the model never does portfolio arithmetic.

03

It refuses to sell unearned alpha. When its own out-of-sample validation finds no edge in a strategy, it says exactly that.

Paper-trades autonomously inside hard guardrails: position caps, daily order caps, a drawdown circuit-breaker, a full audit trail, and a kill switch.

How it works

Link accounts. Model. Advise.

01

Feed it real data

Point it at free market data (CoinGecko crypto, Polymarket prediction markets, SEC EDGAR filings) with no API key. One free Alpaca key adds real equity data and paper execution.

02

Let the desk work

The analyst team researches, debates, and proposes trades while the quant engine computes every number and stamps whether each price was real, cached, or simulated.

03

Judge it on the record

Review the finished research, the paper fills, and the forward-scored track record. Nothing touches real money without you.

Screenshots

Product tour coming soon.

Portfolio dashboard with allocation and benchmark comparison

Product capture coming: Portfolio dashboard with allocation and benchmark comparison.

Multi-agent research run with bull/bear debate and judge verdict

Product capture coming: Multi-agent research run with bull/bear debate and judge verdict.

Track record view with forward-scored agent decisions

Product capture coming: Track record view with forward-scored agent decisions.

Proof and access

What's proven today.

43automated tests green, including live-model runs
5analyst agents debating every research run
3independent switches required before real money

Proof points

  • Runs its full 5-analyst research pipeline end to end on a local Qwen3.6-35B model. No cloud LLM required.
  • 43 automated tests green, including live-model research runs, cross-workspace isolation, paper fills, and the real-money guardrail.
  • Its out-of-sample backtest validation, with statistical significance tests against buy-and-hold, reported no demonstrated edge on a real crypto momentum strategy. The product reports that instead of hiding it.
  • Every price carries provenance (real, cached, or simulated) and every simulated fill is flagged, a discipline born from finding and fixing a real silent-synthetic-fill bug.
  • Real money would require three independent switches armed plus real broker credentials. It can never silently escalate from paper to live.

Pricing and access

Private beta, carefully. The research desk is feature-complete and tested on our own hardware, and we're bringing the first outside seats on slowly. Live trading stays disabled until our securities compliance review is done.

Analytics and research, not investment advice. It never touches real money without you.

Shipping log

What just shipped.

  • June 2026: real market data landed with provenance stamping (CoinGecko, Polymarket, SEC EDGAR) plus Alpaca equity data and paper execution

FAQ

Plain answers.

Is this financial advice?

No. Analytics and research, not investment advice. It never touches real money without you, and live trading is deliberately disabled until our securities compliance review is complete.

Can it beat the market?

We won't claim that. Wealth-Ops ships with a validation harness that runs out-of-sample backtests with significance tests, and when a strategy shows no demonstrated edge it says exactly that. That honesty is the point.

Does it trade with real money?

Not today, and never silently. It separates three venues: a sovereign paper engine, Alpaca paper (real API, fake money), and Alpaca live. Real money would require three independent switches armed plus real broker credentials, and live mode has never been armed.

What does it need to run?

Your own hardware and free data. The full pipeline runs on a local Qwen3.6-35B model, and crypto, prediction-market, and SEC filing data come in with no API key. One free Alpaca key adds real equity quotes and paper execution.

How is this different from other AI trading tools?

The model never does the math. A numpy/pandas/scipy engine computes every number, the agents cite the metrics they used, and every decision is logged at the price it was made and auto-scored for forward-return accuracy. You judge it on a measured record, not on marketing.