14 Unique Hyperliquid Trading Agents. $1,000 Each. Full Transparency.

We just funded 14 Senpi trading agents with $1,000 each and let them loose on Hyperliquid. Every trade, every P&L, every win and every loss - tracked in real-time on a public dashboard anyone can watch.

No cherry-picking. No "backtested results." No "past performance" disclaimers hiding the fact that you're only seeing the highlight reel. This is live capital, live markets, live results. Some are winning. Some are losing. We're showing all of it.

Every agent is open source. Every one links directly to the GitHub repository where the code lives. Any Senpi user can install any of these on their own agent today.

The Senpi Predators

We call them Senpi Predators. Each one hunts differently.

🦊 Fox is the current leader at +18.1% ROI across 50 trades. It catches First Jumps - assets rocketing from obscurity into the top ranks on Hyperliquid's leaderboard in a single scan. The edge is speed: enter on the first signal, before the crowd confirms it. 85% directional accuracy on strong signals. The losses that did happen came from stops that were too tight, not from picking the wrong direction.

🐍 Viper is second at +6.2% ROI with $84 in unrealized profit. It trades range-bound markets - entering at support and resistance when Bollinger Bands are tight, ATR is low, and volume is declining. When every momentum strategy in the zoo is sitting idle because nothing is trending, Viper is working. The chop predator.

🦊 Feral Fox is a trading strategy running on top of Fox - same scanner, higher conviction filters. Score 7+, 3 reasons minimum, regime enforcement, structural invalidation exits instead of time-based cuts. Up 0.6% across 159 trades. The high trade count reflects its earlier, more aggressive configuration; it recently upgraded to a tighter v2 spec.

🐺 Dire Wolf replaces Wolf's configuration entirely with Sniper Mode - FIRST_JUMP only, zero rotation, maker orders on every entry, and DSL High Water Mode for infinite trailing at 85% of peak. Just deployed, scanning for its first high-conviction entry.

🦅 Hawk circles above four high-liquidity markets (BTC, ETH, SOL, HYPE) scanning every 30 seconds. Six filters run on every scan: 5-minute momentum, 15-minute agreement, 1-hour trend, volume spike, chop detection, and funding alignment. Smart money hard block - if the whales are short, Hawk will never go long. Just redeployed after a full architecture overhaul.

👻 Ghost Fox combines Feral Fox's high-conviction entries with DSL High Water Mode. Same First Jump signals, same entry filters, but winners are trailed at 85% of the peak with no ceiling. A First Jump that runs +300% ROE locks +255% instead of capping at a fixed tier. The upgrade path for anyone already profitable on Feral Fox. Just deployed.

🦂 Scorpion is unlike anything else in the zoo. It doesn't make its own trading decisions. It discovers top-performing whales from the leaderboard, monitors their actual position changes, and mirrors entries when multiple whales align on the same trade. A 10-minute delay filter prevents copying noise. Per-whale performance tracking automatically demotes whales that stop performing. The sting: when a tracked whale exits, Scorpion exits immediately. Just deployed.

🐍 Mamba is a trading strategy on Viper - same range-bound entries but with DSL High Water Mode. Most range trades bounce for small profits and both Viper and Mamba capture those equally. But about 1 in 5 range entries catches a breakout where price escapes the range and trends. Standard Viper caps that gain. Mamba rides it. Just deployed.

🐍 Cobra enters only when three independent signals converge: multi-timeframe price momentum (5m, 15m, 1h all agreeing), volume confirmation (real buying pressure, not noise), and open interest growth (new money entering, not just repositioning). Triple gate means most scans produce nothing. When Cobra strikes, all three signals confirm. Down $7 on 2 trades - early days.

🦈 Shark follows smart money consensus - when 3 or more of the top 10 traders on the leaderboard are aligned on the same asset and direction. A secondary pipeline estimates where liquidation cascade zones sit and enters just before price reaches them. Down $18.43 on 7 trades. Shark is built for volatile, cascading markets - the current chop isn't producing the setups it needs.

🦬 Bison is the conviction holder. Top 10 assets by volume only. Enters when 4h trend structure, 1h momentum, and smart money all converge. Holds through pullbacks with the widest DSL bands in the ecosystem - a trade can retrace 25% ROE without getting stopped. Re-evaluates the thesis every 5 minutes and exits when the conviction breaks, not when price retraces. Down $30.55 on 18 trades. Bison needs multi-day trends to play out; in choppy markets it loses on structural invalidation exits.

🐊 Croc trades funding rate arbitrage - entering against extreme funding to collect the rate while positioning for the mean-reversion snap. Down $59.91 on 107 trades. The high trade count tells the story: the current funding environment isn't producing the extreme divergences Croc needs. This is a strategy that performs best in volatile, one-sided markets where funding rates go parabolic.

🐅 Tiger is the most complex system in the zoo. Five parallel scanners - compression, correlation, momentum, funding, and reversion - all feeding into a prescreener that narrows 230 assets to the top 30. A meta-optimizer called ROAR tracks which scanners are producing winners and automatically throttles the underperformers. Down $68.17 on 37 trades. Tiger's breadth is both its strength and its weakness: five scanners means five sources of potential entries, and in a directionless market, that can mean over-trading.

Skills vs Trading Strategies

Some of these are unique skills - standalone agents with their own scanners and signal logic. Fox, Viper, Hawk, Shark, Scorpion, Cobra, Bison, Croc, and Tiger each have their own codebase.

Others are trading strategies - configuration overrides on top of an existing skill. Feral Fox, Ghost Fox, Dire Wolf, and Mamba run the same underlying scanner as their parent skill but with different entry filters, DSL settings, and risk parameters. Same code, different personality.

The distinction matters because it shows how the system scales. Writing a new skill requires building a scanner, which takes real development time. Creating a new trading strategy requires changing a JSON config file, which takes minutes. Most users will never write a skill. They'll deploy Fox or Viper with defaults, and over time their agent will help them tune the variables based on real performance data - creating a personalized trading strategy in the process.

This is the three-layer architecture we call SAITA (Senpi AI Trading Architecture):

Plugins are shared infrastructure - trailing stops, risk management, fee optimization, health checks. Maintained once, every skill benefits. When we fix a bug in the trailing stop engine, all 14 agents get the fix.

Skills are the unique trading logic - the scanner that embodies a thesis about how to make money. Fox's First Jump detector. Viper's range-bound analysis. Scorpion's whale tracking. Each skill is a thin layer on top of the shared plugins.

Trading Strategies are saved configurations - the specific numbers that tune how aggressively a skill behaves. Ghost Fox is Fox's skill with Feral's entry filters and High Water trailing. Mamba is Viper's skill with wider DSL bands. The skill is the predator. The trading strategy is how you teach it to hunt.

What We're Learning in Public

Three things are becoming clear from watching 14 agents trade live:

Leaderboard momentum is a real edge. Fox is up 18.1% catching First Jumps - assets that explode from outside the top 20 into the top 10 in a single scan. The signal has 85% directional accuracy on high-score entries. The losses came from tight stops killing correct-direction trades too early, which is why every subsequent Fox variant (Feral, Ghost) has progressively wider stops and fewer time-based exits.

Range-bound trading works when nothing else does. Viper is quietly profitable in a market where most momentum strategies are flat or losing. When BTC chops sideways for two days, Fox and Dire Wolf scan and find nothing. Viper finds support/resistance entries every few hours. The zoo benefits from having strategies that are anti-correlated - when one type is idle, another is hunting.

Trade frequency is the silent killer. The pattern across all 14 agents: the ones losing most are trading most. Croc has 107 trades and is down $60. Tiger has 37 trades and is down $68. Fox has 50 trades and is up $181. The fee math is unforgiving on Hyperliquid - every round-trip at 10x leverage costs real money. Every upgrade we've shipped across the zoo has tightened the same thing: fewer trades, higher conviction, wider stops.

DSL High Water Mode

The newest innovation across the zoo is something we call DSL High Water Mode. Traditional trailing stops lock fixed ROE amounts: "at +10% ROE, lock +5% ROE." This creates a ceiling - the locked amount is static no matter how far the trade runs.

High Water Mode locks a percentage of the peak instead. At Tier 4 (+20% ROE), the stop is always 85% of wherever the high-water mark is. If the trade runs to +100% ROE, the stop is at +85%. If it runs to +500%, the stop is at +425%. No ceiling. The geometry stays constant.

The full spec is open source and any skill can adopt it as a drop-in configuration.

What Comes Next

All of these agents are beta. We're actively tuning configurations, fixing bugs discovered in production, and shipping upgrades - often multiple times per day. The public dashboard will reflect every change in real-time.

Behind the scenes, we're building SAITA into a complete platform. The plugin infrastructure will make deploying new strategies a couple minutes of work and writing new skills as easy as an hour-long yaml file. Shared plugins mean every skill benefits from improvements to trailing stops, risk management, and fee optimization without any per-skill work.

We're also adding more predators. Eagle (correlation breaks and macro events), Panther (breakout scalping), Wolf Nightshift (Asian session momentum), OWL (contrarian crowding unwinds), OWL Aggressive, and Tiger Sniper are all in various stages of development and testing. The zoo is growing.

Check back regularly. The leaderboard shifts daily, and every number on it is real.

The Home of Hyperliquid Agents.