Describe it. Approve it. It trades.
Tell DXAP your strategy in plain words. We turn it into an agent, you set the limits and approve it, and it trades Hyperliquid around the clock — inside your rules, never outside them. Built by DXRG.ai, an agentic trading research lab.
From your idea to a live agent.
Describe an idea, set how much risk the agent can take, and approve it for paper or live trading. Agents run on our servers, so closing the browser does not stop them. The agents below are real and anonymized, shown from the prompt the user wrote to the trades the agent made.
Live numbers from Pre-Alpha agents. Past performance doesn't guarantee future results.
HIGH-RISK LONG SHOT. I WANT AROUND 50% IN 8 HOURS AND I ACCEPT A REAL CHANCE OF A BIG LOSS. PILE INTO THE SINGLE HIGHEST-MOMENTUM, HIGHEST-VOLATILITY LIQUID NAME ON A STRONG BREAKOUT WITH EXPANDING VOLUME. 8-15X, BUT ALWAYS PRE-COMMIT A HARD STOP AND SCALE OUT INTO THE SPIKES, THEN TRAIL A RUNNER. PRESS WINNERS, CUT LOSERS INSTANTLY. IF THERE'S NO SETUP, WAIT. SWING FOR THE FENCES.
Breakouts with real volume tend to keep going. Concentrate into the single best one with size, scale out into the spikes, and cut instantly when it fails. When nothing qualifies, hold cash.
- Side
- Long
- Size
- $65k
- Leverage
- 10x
- Result
- +$3,162
Scaled out of the parabolic leg, then trailed and closed the runner.
- Side
- Long
- Size
- $82k
- Leverage
- 10x
- Result
- -$244
Breakout did not follow through, so it cut the position fast and small.
- Side
- Long
- Size
- $75k
- Leverage
- 10x
- Result
- -$167
No continuation after entry, closed it small and moved on.
- Day 1Held flat by design
Nothing cleared its bar (a 1h move over 2%, expanding open interest, and over $50M of volume), so it stayed in cash rather than force a trade.
- Day 4Passed on a fake-out
JUP looked like a breakout, but the 1h move flipped negative and volume was too thin to size into, so it skipped it.
- Day 8Pressed the winner
XPL broke out cleanly. It went in at 10x, scaled out into the spike, and trailed the rest.
- Day 12Cut the losers fast
SOL and ZEC failed to follow through after entry. It closed both small and went back to cash.
A research desk behind every turn.
A DXAP agent is not a chatbot with a price feed. Each turn it can call the tools a trading desk would reach for, then act inside your limits. The capability list grows by the day — what the lab learns ships back into the harness.
- SubagentsSpecialist workers — research, news, and technical analysis — run inside the harness and feed the decision. A desk of agents, each scoped to one job.
- X watchThe agent follows curated X accounts you approve, so positioning and sentiment shifts land on the turn they matter — not in a weekly digest.
- Prediction marketsPolymarket odds sit next to price as an assessment input. A thesis gets checked against what markets are paying for, not just what charts did.
- TriggersStanding triggers fire extra turns when conditions hit — price levels, volatility, or your own rules. The agent is awake for the move, not after it.
- Self-schedulingWhen something needs watching, the agent sets its own follow-up wake-ups and says why.
- On the recordEvery turn is written down — data in, decision out, including "no trade". You can audit any turn after the fact.
The harness is the edge.
A trading agent is a model plus a harness. General harnesses such as Claude Code or Pi are built for writing code, not for trading. DXAP's harness is built for markets. On every turn the agent gets fresh market data, its live portfolio, and its strategy. Hard risk limits sit outside the model, so the agent cannot override them. Agents wake on schedules and on market triggers, around the clock. In our replay benchmark, the same models returned about 22% inside DXAP and lost money in the other setups. DXAP is a lab product: every turn agents run on the platform feeds our benchmarks, and what we learn ships back into the harness. The system you trade on is the experiment.
Read the researchMedian return on our internal replay benchmark. Same models and same market data in each setup. Replay benchmarks don't guarantee live results.
DXAP is in Pre-Alpha. We ship changes by the day — expect things to move.
Join the waitlist to get a referral code. If you already have one, enter the app and connect your wallet.
Enter app