Private trading agents · By application
Your own trading agent,
built to your mandate.
We design and build individual investors a private, multi-agent trading system — researched, backtested, and connected to your brokerage. It proposes; you approve every trade.
By application. We take a limited number of builds.
The problem
Building a real trading agent is a full-time engineering job.
Data pipelines. Backtesting that isn't quietly lying to you. Risk management. Model orchestration that stays reliable when markets move. Broker integration. Most people who want an automated agent never ship one — and the ones who do often can't tell whether the results are skill or luck.
What you get
A private agent, built to your goals
Not a generic bot. We design the strategy around your objectives, risk tolerance, and capital.
Multi-agent research
Analysts, opposing researchers, and a risk committee cross-examine every decision before it reaches you.
Backtested before a dollar moves
Validated over multi-year history and paper-traded first — with the losing years shown to you, not hidden.
You approve every trade
Nothing reaches your brokerage until you approve it. Pause or disable the agent at any time.
How it works
Apply
Tell us your goals, capital, and brokerage. We review fit and decline what we can't build well.
We design and build
We architect, backtest, and paper-trade your agent until the evidence is good enough to show you.
Your agent runs
It surfaces trades for your approval, places the ones you greenlight, and reports back.
Under the hood
Real engineering, and the evidence to go with it.
- Multi-agent research pickers
- Rule-based strategy composites (MACD/RSI, moving-average crosses, oscillators)
- Portfolio and risk managers translate signals into orders
- Multi-year backtest engine with paper-trading runners
- Options and income strategies (the Wheel, covered calls)
Why multi-agent, and what it costs
We tested a single-agent picker against the multi-agent design across eight years of market history. Here is what we found — including the years both of them lost.
Across eight years, the multi-agent design beat the S&P in six of eight — and never had a losing year. The single-agent version managed three.
Annual return vs S&P 500
Multi Single
Vertical marker = S&P 500 that year. Both strategies lost to it in 2019 and 2025 — neither adds value when everything rises together.
| Year | Single-agent | Multi-agent | S&P 500 |
|---|---|---|---|
| 2018 | -2.8% | 17.2% | 2.4% |
| 2019 | 15.2% | 14.1% | 30.9% |
| 2020 | 4.7% | 38.3% | 15.9% |
| 2021 | -12.6% | 41.5% | 21.3% |
| 2022 | 13.1% | 21.5% | -14.7% |
| 2023 | 58.4% | 45.0% | 30.7% |
| 2024 | 31.6% | 44.7% | 28.7% |
| 2025 | 2.5% | 4.8% | 19.6% |
The trade-off: a multi-agent decision runs 12 model calls — four analysts, opposing researchers, a risk committee, a portfolio manager — against one. Roughly 12× the compute, for roughly twice the return at lower volatility.
Backtest, not live trading.Simulated on 2018–2025 history with a 10 bps round-trip cost model; no real money was traded, and simulated results do not predict future results. Returns are shown before tax. Our own diagnostics put a ±5–25pp execution-luck band on these figures. Max drawdown on the related single-model baseline was −42.2%, deeper than the S&P's −33.2%.
Who it's for
Individual investors who want a serious, systematic trading agent — and would rather have engineers build it than spend a year learning to.
You bring the capital and the mandate; we bring the system. It is not a get-rich scheme, it will not beat every market, and it is not for money you can't afford to put at risk. If that reads as a warning rather than a pitch, you're the sort of client we work well with.
Apply for the waitlist
Tell us about your goals. We'll be in touch about building your agent.