Founder access · closed beta

Autonomous AI
trading agents

The discipline you lose in the market — running while you sleep.

Soulquant gives you an autonomous agent that runs your strategy on your exchange account. Patient, rule-bound, and — by design — mostly doing nothing. You bring the strategy and the keys. We bring the infrastructure.

Claim a founder seat → See how it works
Starts in paper mode Your keys never leave your vault Not a signal service
ETH-PERP
agent
LIVE
Current state
state
In-trade short
bias
short
regime
Down
trailing stop
target 1,896
next wakeup
in 47m · or price trigger
today
14 wakeups · 0 trades
Run note
Price extended into resistance after a climactic push. Setup not clean. Holding idle until structure confirms. Doing nothing is the expected outcome.
Built for crypto perpetuals You bring your exchange keys You own your strategy It trades only when it should
The real problem

Your strategy isn't the problem. Your discipline is.

Most traders don't lose because they lack an edge. They lose because they can't execute their own rules at 3am, after three losses, when the chart is screaming.

You overtrade
Boredom and FOMO turn a patient plan into ten impulsive entries. The best trade is usually no trade — but you take it anyway.
You chase
You enter after the move is already extended, right where a disciplined trader would wait — or fade. Every loss has the same fingerprint.
You let losers run
Hope widens the stop. A structural invalidation becomes a liquidation. The one rule that mattered is the one you broke.
How it works

An agent that works like a disciplined human trader

Not a bot firing on every candle. A patient agent that carries a state, wakes only when something it cares about happens, and asks one narrow question at a time.

01
It carries a state, it doesn't re-scan
Idle, Watching, or In-trade. The agent never re-evaluates the market from zero — it only checks whether the one thing it's watching has happened. That's what lets an LLM trade with patience instead of noise.
Idle → Watching → In-trade
02
It wakes on a schedule — and on triggers
Like a human checking in. It sleeps between setups, then wakes on a time cadence or when price crosses a level it set. Most wakeups end in "nothing to do here," and that's the point.
time + price triggers
03
It runs your risk profile & aggressiveness
Two independent dials, set at onboarding: how loosely it promotes a setup, and how much it sizes when it does. The agent reads and judges; deterministic code handles sizing, leverage caps, and risk gates.
judgment + hard gates
04
It learns your asset, and that self compounds
After every trade it logs an observation. A daily review distills those into a curated playbook. Over weeks, your ETH agent and your SOL agent genuinely diverge — each learns how its own asset moves under your strategy.
a bounded, growing edge
The separation that protects you

Judgment and execution never sit in the same hands

The agent that decides whether to trade can't size or place orders. A separate controller owns risk — with hard gates the agent never sees.

The Promoter only signals
Side, entry, stop, target, confidence — never an amount. It outputs a signal and stops there.
The Controller sizes & gates
It reads your live balance, computes position size from risk × confidence, and rejects deterministically before any order if a risk gate fails.
Liquidation is never the stop
The stop is structural invalidation, sized so price hits your stop long before liquidation. Leverage is a sizing tool, never a thesis.
agent.run — eth-perp
wakeup 14:32 UTC · price trigger @ 2,011
state Watching short
regime Down · trend intact
read extended into resistance, not clean
decision hold — no setup
wakeup 18:05 UTC · rejection confirmed
promote short · entry 1,994 · SL 2,022 · TP 1,896
confidence 7/10
controller risk ok · size 0.42 · lev 4x
EXECUTED short opened · stop << liquidation
Strategy · your doctrine
EDITABLE
## Coin ETH-PERP
## Direction trend-aligned only
## Aggressiveness measured
## Setups
• pullback-continuation
• breakout-retest
## Session rules pause weekends
## Confidence rubric 1–10
Oracle — all setups defined, regime & timeframes present, direction policy consistent. Ready to finalize.
Your strategy. Your keys.

We're the infrastructure — not your fund manager

Soulquant never holds your money. You connect your own exchange account with trade-only keys, and you own the doctrine the agent runs.

Bring your own keys, encrypted
Trade-only API keys, encrypted at rest in a managed vault, decrypted only at run time. Never withdrawal scope. Never logged.
Start from a template, or write your own
Begin with a proven strategy template, or draft your own doctrine in plain language. A guided interview checks it for consistency before it goes live.
Total transparency, versioned
Every run, trade, and learning is stamped with the exact instructions that produced it. Ask "what did my agent do on the 27th, and why?" — and get a precise answer.
Built to be wrong safely

The agent is wrong sometimes. The architecture survives it.

We don't design for an oracle that never misses. We design for small losses, capped exposure, and rules that can't be overridden in the moment.

Invariants over everything
Leverage caps, stop discipline, regime gates, never counter-trend. A "lesson" that wants to relax a hard rule is flagged, not followed.
Paper mode by default
Every new agent starts in simulation. Going live is a deliberate, confirmed switch — never an accident, and never before you've watched it behave.
Patience is structural
The system optimizes for trade quality, not frequency. Every component makes "no trade" the easier path. It's built to free your attention, not pull it.
Radical honesty
The whole design is legible: every decision, gate, and rejection is recorded with the reasoning that produced it. Nothing about your agent is a black box.
Radical honesty

What Soulquant is not

The fastest way to spot a scam in this space is what it refuses to admit. So here's ours, up front.

Not a get-rich machine
No performance promises. It chases the best risk-adjusted return achievable without your attention — not the biggest number.
Not a signal service
It trades your own capital, autonomously. There's no group chat selling you calls.
Not a scalper
No 1m/5m reaction game. Edge lives in higher-timeframe structure, where an LLM can actually read it.
Not an oracle
It will be wrong. The whole design is about being robust to that — small, capped, invariant-protected.
Pricing

Pay for infrastructure, not promises

A simple platform seat, plus a prepaid balance for the compute your agents actually use. You only pay when your agents run — and patient agents run rarely.

Founder seat · first 100 only
Founder
Locked for life. The price goes up as seats fill.
$39/ month
then $79 → $149 as the community grows
Run multiple autonomous agents in parallel
Strategy templates + build-your-own doctrine
Backtest harness to vet a system before it's live
Founder pricing locked permanently
Claim your seat
Compute wallet
Prepaid balance for agent runs. Auto-reload optional.
Usage
top up from $20 — you pay for tokens your agents use, nothing when they sleep
balance $32.40 / $50.00
auto-reload on+$50 at $10
Transparent token-level metering per run
Auto-reload via Stripe with daily caps
Low balance? Agent pauses — it never trades broke
Join the waitlist
FAQ

The questions a serious trader asks

Do you ever touch my funds?
No. Soulquant never holds, custodies, or moves your money. You connect your own exchange account with trade-only API keys — no withdrawal scope, ever. Keys are encrypted at rest and decrypted only at run time.
Will it make me money?
We make no performance promises, and you should distrust anyone who does. The agent chases the best risk-adjusted return achievable without your attention — small, capped, invariant-protected. It will be wrong sometimes; the architecture is built to survive that.
Which exchanges do you support?
We focus on crypto perpetuals. Support launches with the major perps venues and expands over the beta — you connect via trade-only API keys on the exchange you already use.
Do I need to know how to code?
No. You write your doctrine in plain language, or start from a proven template. A guided interview checks it for consistency before it goes live. Everything the agent does is legible without reading a line of code.
What does it cost to run?
A founder seat is $39/month, locked for life for the first 100. Beyond that you prepay a compute wallet and pay only for the tokens your agents actually use — and patient agents run rarely. If your balance runs low, the agent pauses; it never trades broke.
Founder access · closed beta

Give your strategy a soul.

The first 100 founders lock in $39/month for life and start in paper mode. Be early — the price only goes up from here.

Paper mode first · trade-only keys · no performance promises, ever.