Plutus
Cursor for trading strategies · private beta

Trade your instinct.
Put it on rails.

Say a setup in one sentence. Plutus turns it into a real strategy, tests it honestly against two years of tape, and runs it from your own wallet under hard risk limits.

No code · self-custody · we can never withdraw
plutus://desk/liveHyperliquid
you ▸
compiled spec
entry:funding_3d < 0 && oi_4h > +3%
exit:tp 5.0% · sl 2.0%
size:long · 3x · 4h
backtest equity · 2yr
hypothetical · gross of fees
the gate · 7 checks
✓sharpe1.31
✓max drawdown−11.4%
✓trades63
✓walk-forward3 / 3
✓overfitting0.22
✓significancep 0.03
✓data quality99.4%
CLEARS THE GATE · PAPER NEXT21 days on the live tape, no money at risk, before it can be deployed

231,000 traders. Three bad options.

hyperliquid active traders, mar 2026
1

Trade it by hand

Discipline breaks. You miss the entry, move the stop, revenge-trade at 3am.

2

Buy a black-box bot

Hand your keys to a stranger whose only evidence is a backtest screenshot.

3

Become a quant

Data, backtester, execution, risk engine. A team and months of work.

74 to 89% of retail derivatives accounts lose money. Not because the instincts are worthless, but because there is no path from instinct to disciplined execution that does not require becoming an engineer.

idea → proof → live → learn, then again

The loop that never stops.

One evaluator runs backtest, paper and live, so what you tested is what trades. We will not let you fool yourself: a significance test asks whether your edge could just be luck.

monitor feeds the next idea
01 / 06Describeplain English in↻ then again
1Describe
2Backtest
3Gate
4Paper
5Live
6Monitor
one risk engine behind every connector

One engine. Venue is a field.

A new venue is a connector behind the same risk engine, not a second product. One is live today and the rest are honestly labelled.

trading livebuilt, not liveplanned
PlutusRISK ENGINEHyperliquidliveLighterbuiltPolymarketbuiltArcusplannedKalshiplanned

Live to backtest right now.

real stored history, not a promise
BTC
2yr history85k bars
1m1h4h
ETH
2yr history85k bars
1m1h4h
SOL
2yr history85k bars
1m1h4h
Prediction marketsnot backtestable yetBinary payoffs need their own engine. Not started.

Your keys. Your coins. Always.

the agent can trade, never withdraw
Youwallet · keys
approve once
Trade-only agentplace · cancel orders✕ cannot withdraw
orders only
Venuehyperliquid

You sign an EIP-712 approval from your own wallet. The agent can place and cancel orders. It cannot withdraw, and it cannot approve other agents.

Be wrong. Don't blow up.

circuit breakers checked before every live entry
10×leverage cap

Orders above it never reach the venue.

70%max margin used

Over-margin entries are skipped.

5%daily loss stop

Hit it and entries pause till tomorrow.

BLOCKS
15%drawdown kill

Pauses every strategy you run.

Enforced by the engine. A bad idea costs a stop, never your account.

Every fill public. Wins and losses.

illustrative preview · opens with the beta
Illustrativeexample rows, not live strategies
strategytrader7dsharpelive pnl
1BTC funding-flip · 3x0xa1…f5e1.31+6.2%
2ETH mean-revert · 2x0x6f…44d0.34+0.4%
3SOL breakout · 2x0x88…088-0.71-4.8%

Sample layout, not real results. When the Arena opens, every number here is a verifiable on-chain fill, including the losing ones.

Questions traders ask.

No. You describe the setup in plain language and Plutus compiles it into typed, checkable rules you can read and edit before anything runs.

In your own venue wallet at all times. You sign an EIP-712 approval that lets our agent key place and cancel orders. It cannot withdraw, and it cannot approve other agents.

Hyperliquid perps are live end to end. Lighter has a connector built but no live fill yet. Polymarket has an executor but no backtest engine for binary payoffs, so it is not tradeable here yet. Arcus and Kalshi are planned.

Seven checks on a two-year backtest: Sharpe, drawdown, sample size, walk-forward, an overfitting score, a bootstrap significance test, and data quality. One failure stops it. Then paper trading against the live tape, 21 days by default, which you can extend.

We do not promise returns and we will not show you a number that implies one. Our own live strategies are net-negative after fees to date, and they are internal demos we deployed before the gate became a deployment precondition. That result is the reason the gate exists and the reason we publish what we predicted against what happened.

Funding is modeled. Trading fees and slippage are not, yet, so every backtest figure on this site is gross and labelled hypothetical. Costing them properly is the current piece of work, because fees are exactly what turned our own best-looking strategy into a loser.

The health score drops, new entries pause automatically, and the drift becomes suggested rule changes you can replay before resizing.

Your instinct. Systematized.

Private beta is opening in waves. Get on the list and we'll reach out before your slot.

Plutus
© 2026 Plutus · self-custody · no funds held · deck