Prebuilt Strategies
The eight ready-made templates on /create, what each one trades, the exact decision loop it runs, and every setting with its default and allowed range.
When you create an agent at /create, the first step is choosing a strategy. Eight templates are prebuilt; a ninth, "Build your own", is covered in Custom Strategy Builder. Every number on this page comes from the rules the engine enforces (src/lib/agents/rules.ts and src/lib/agents/engines.ts): a setting outside its range is rejected, and the agent keeps trading on its last valid rule set.
The templates at a glance
| Template | What it trades | Market data | House agents that launched on it |
|---|---|---|---|
| Uniswap V4 launches | New Robinhood Chain tokens launched on their own hooked Uniswap V4 pool | Neuron indexer | Quantara |
| Stablecoin yield | USDC in blue-chip lending pools on Ethereum, Base and Arbitrum | DefiLlama APYs, live gas | Yield Yoda, Kai |
| Hedged LP farming | BNB Chain pools pairing a stablecoin with BNB, BTC or ETH, optionally hedged | DefiLlama APYs, Hyperliquid prices and funding | Nexi, El Toro, VoltXBT, Sola |
| Buy-the-dip DCA | BTC and ETH (and optionally XMR), bought in tranches | Hyperliquid prices | Ryuu, Monero Chan |
| Majors momentum | The strongest of 11 ETH-ecosystem majors | Hyperliquid prices | None yet |
| Perp trend following | BTC and ETH perpetuals, long or short | Hyperliquid prices and funding | Kage |
| Funding-rate carry | Spot long plus an equal perp short on the best-paying markets | Hyperliquid funding and volume | Vera Alpha, DeFi Purr |
| Blue-chip basket | A fixed basket of six large caps | Hyperliquid prices | None yet |
House agents can change rules after an approved loss review, so an agent's current settings can differ from the template defaults below. Its profile at /agents/<id> always shows the rules it trades on now.
What every template has in common
- Paper trading on live data. On testnet, no real orders are placed. The agent worker runs every agent every 30 seconds, decides on live market data, and reports the result on-chain as a return in basis points. See Public PnL & Trade History.
- Real costs on every simulated trade. Spot trades pay 0.10% of the amount traded (DEX fee plus slippage on majors at these sizes). Perp fills pay Hyperliquid's 0.045% taker fee. Gas is charged at the chain's live gas price: 180,000 gas per swap, 250,000 per lending deposit or withdrawal, 400,000 to add LP and 300,000 to remove it.
- Reporting cadence. Trades are reported as soon as they happen. Plain marks go out every 15 minutes for launch-trading agents and every hour for the other strategies.
- New rules act immediately. When an agent's settings change, it scans right away instead of waiting for its next scheduled rebalance.
- Switching strategy closes the old book. If an approved change moves an engine agent to another engine, the old positions are closed first, paying their exit fees and withdrawal gas.
Uniswap V4 launches
What it trades: freshly launched Robinhood Chain tokens that went straight onto a Uniswap V4 pool with a hook deployed for that token. The data comes from the Neuron indexer, which follows Robinhood Chain mainnet; the agent paper-trades those real tokens and reports to its testnet vault.
How it decides, every tick:
- 1Ask the indexer for tokens passing all filters, ranked by 1h volume, highest first.
- 2Drop any pick whose round trip (fees, creator tax, price impact at this position size, gas) would cost more than the max round-trip cost.
- 3If it holds a token: sell at take profit, sell at stop loss, otherwise keep it while it ranks in the top 3 or while the 30-minute minimum hold has not passed. After that, rotate.
- 4With no position (or after selling), buy the best remaining pick. If guard tools are connected, up to three picks are checked in order and a vetoed one is skipped.
Default filters:
| Filter | Default | Why |
|---|---|---|
launchpad | is uniswap_v4 | Only tokens launched directly on V4 |
hooked | is true | Pool has a hook of its own |
round_trip_back | at least 0.9 | A simulated 0.01 ETH buy and immediate sell returns at least 90% (no honeypot hook) |
buyers | at least 8 | More than a handful of distinct buyers |
buy_sell_ratio | above 2 | Buyers have put in over twice what sellers took out |
top_maker_pct | below 60 | No single buyer took most of the supply bought |
mc | above $8,000 and below $200,000 | Small, but not empty |
Position rules (shared with every token-filter agent):
| Setting | Default | Range | Meaning |
|---|---|---|---|
| Position size | 5% | 5% to 50% | Share of the vault put into one token |
| Take profit | +100% | +5% to +500% | Sell once up this much from entry |
| Stop loss | -40% | -90% to -2% | Sell once down this much from entry |
| Hold while in top | 3 | 0 to 10 | Keep the token while it ranks in the top N picks |
| Min hold | 30 min | 0 to 240 min | No rotation before this (take profit and stop loss still apply) |
| Max round-trip cost | 12% | 1% to 30% | Skip tokens costing more than this to get in and out |
Specific risks: most hooked V4 launches cannot be sold at a fair price, which is why the round-trip filter exists; but a hook owner can change behavior after the last check, and checks stop one hour after the first trade. Liquidity is thin, so price impact is real even at 5% of the vault. A rug can move more than one report allows; the excess is carried into the next reports rather than dropped.
Stablecoin yield
What it trades: USDC lent in single-asset pools on Aave v3, Compound v3, Morpho Blue, Spark Savings, Fluid Lending or Euler v2, on Ethereum, Base or Arbitrum, at their live DefiLlama APYs.
How it decides:
- 1Every tick, each held pool accrues its live APY since the last mark.
- 2On each rebalance, it lists eligible pools at or above the min pool size, sorted by APY.
- 3Each pick gets the smaller of "max per pool" and one divided by "pools at once". With the defaults that is 33.3% each, fully invested.
- 4A new pool is skipped when one rebalance period of its yield on that share would not cover twice the gas (a deposit now and a withdrawal later).
- 5A pool it already holds stays unless a replacement pays at least the min APY gain more.
| Setting | Default | Range | Meaning |
|---|---|---|---|
| Min pool size | $100M | $10M to $2,000M | Only pools with at least this much deposited |
| Pools at once | 3 | 1 to 5 | How many pools to split across |
| Max per pool | 40% | 20% to 100% | Largest share of the vault in one pool |
| Rebalance every | 7 days | 1 to 30 days | How often to move toward the best pools |
| Min APY gain to move | 0.5% | 0% to 5% | Stay put unless a pool pays this much more |
Specific risks: APYs are variable and can fall right after a move. If "max per pool" times "pools at once" is under 100%, the rest sits in cash earning nothing (for example 1 pool at 40% leaves 60% idle). On Ethereum, gas can keep a small vault out of pools entirely.
Hedged LP farming
What it trades: liquidity in the deepest BNB Chain pools that pair one stablecoin (USDT, USDC, FDUSD, DAI and others) with BNB, BTC or ETH, earning the pool's live APY. With the hedge on, it shorts half of the volatile leg on Hyperliquid.
How it decides:
- 1Every tick, each position is marked as a full-range LP: its value moves with the square root of the volatile coin's price change, times the APY accrued. With the hedge, the short on half the volatile leg offsets the first-order price move and collects (or pays) half the hourly funding rate.
- 2On each scan, candidate pools are ranked by APY. A pool it already holds gets the min APY gain added, so moving has to pay for itself.
- 3It farms the top pools at equal weight. Entering or leaving costs 0.05% (swapping half into the volatile leg) plus 0.0225% for the hedge, plus LP add and remove gas.
| Setting | Default | Range | Meaning |
|---|---|---|---|
| Min pool size | $2M | $0.5M to $100M | Only pools with at least this much liquidity |
| Pools at once | 2 | 1 to 3 | How many pools to farm |
| Hedge with a perp short | on | on or off | Short half the volatile leg to offset impermanent loss |
| Scan every | 168 hours (7 days) | 6 to 720 hours | How often to rescan and move |
| Min APY gain to move | 0% | 0% to 50% | Keep a pool unless a replacement pays this much more |
The Multichain LP + Relay tool widens the search to Ethereum, Base and Arbitrum, counts only the trading-fee part of APY (capped at 100%), leaves out vote-escrow DEXs (Aerodrome, Velodrome, Thena), and charges a live Relay bridge quote when capital moves between chains (5 bps if Relay cannot quote).
Specific risks: without the multichain tool, the APY used includes reward tokens, which can be short-lived. The hedge removes first-order price exposure only: large moves in either direction still lose to impermanent loss, and funding can turn against the short. Real concentrated-liquidity pools behave differently from the full-range model.
Buy-the-dip DCA
What it trades: BTC and ETH, plus Monero (XMR) if you give it a share. It buys and holds.
How it decides:
- 1It checks once a day at most, after 14:00 UTC.
- 2With the trend filter on, it first sells any held coin trading more than 1% below its simple average over the trend window, and only buys dips in coins that are not below it.
- 3In red-day mode ("Buy every" at 0), a coin qualifies when it is down at least the red-day threshold over the last 24 hours. With the News sentiment tool connected, it skips a coin whose news scores very negative.
- 4In scheduled mode ("Buy every" above 0), it buys a tranche every N days whatever the market did.
- 5Each buy is the tranche size, capped by the room left under max invested and by the cash on hand, split between the coins by their shares.
| Setting | Default | Range | Meaning |
|---|---|---|---|
| Tranche size | 5% | 1% to 20% | Share of the vault each buy uses |
| Red day | 2% | 0.5% to 10% | Buy only when down at least this much over 24h |
| Max invested | 90% | 10% to 100% | Stop buying past this much invested |
| BTC share | 50% | 0% to 100% | Split of each buy between BTC and ETH |
| Trend filter | 0 days (off) | 0 to 120 days | Buy dips only above the average; sell below it |
| Buy every | 0 days (red days only) | 0 to 14 days | Fixed schedule, red day or not |
| XMR share | 0% | 0% to 100% | Share of each buy that goes to XMR |
Monero Chan launched with a tranche every 3 days, 100% XMR, 5% tranches, up to 100% invested and no trend filter.
Specific risks: with the trend filter off, it never sells, so it rides every drawdown. A long bear market fills the vault up to max invested at falling prices.
Majors momentum
What it trades: ETH, LINK, AAVE, UNI, LDO, ENS, ARB, OP, ENA, ONDO and PENDLE, or stables when nothing is trending.
How it decides:
- 1On each rebalance, it ranks the 11 coins by trailing return over the lookback.
- 2It buys the top N whose return is at least the min momentum and (with the trend filter on) whose price is above its moving average. Each gets 1/N of the vault; unfilled slots stay in stables.
- 3Between rebalances, with the trend filter on, it sells any holding that falls below its average.
| Setting | Default | Range | Meaning |
|---|---|---|---|
| Momentum lookback | 72 hours | 24 to 336 hours | Window used to rank assets |
| Assets held | 2 | 1 to 4 | How many of the strongest to hold |
| Rebalance every | 24 hours | 6 to 168 hours | How often to re-rank |
| Min momentum | 0% | -5% to 20% | Only buy assets up at least this much |
| Trend filter | 0 hours (off) | 0 to 168 hours | Only hold assets above this moving average |
Specific risks: momentum reverses. Fast rebalancing in choppy markets pays fees on every rotation, and with one asset held the vault is fully concentrated.
Perp trend following
What it trades: BTC and ETH perpetuals on Hyperliquid prices, long or short, half the vault as margin per market.
How it decides, per market:
- 1Compute a fast and a slow exponential moving average of hourly closes. Long when fast is at or above slow, short when below.
- 2Inside the flip buffer (the averages closer than the buffer), keep the current side, or stay flat if there is none.
- 3If price has moved against the position by the stop loss (measured before leverage), close it and stay out until the signal flips.
- 4Do not open a side whose funding costs more than the crowding guard a year, or (with the News tool) a long against negative news or a short against positive news. An open position on the other side is closed.
- 5Every fill pays 0.045% times leverage; a flip pays it twice. Funding accrues at the position's leverage.
| Setting | Default | Range | Meaning |
|---|---|---|---|
| Leverage | 2x | 1x to 3x | Position size relative to margin |
| Fast average | 6 hours | 2 to 24 hours | Fast EMA |
| Slow average | 24 hours | 12 to 96 hours | Slow EMA |
| Stop loss | 6% | 2% to 20% | Price move against the position that closes it |
| Flip buffer | 0.3% | 0% to 3% | Minimum gap between averages to change side |
| Crowding guard | 0% (off) | 0% to 100% | Skip a side whose funding costs more than this a year |
Connecting the Hyperliquid positioning tool turns the crowding guard on at 25% if it is off.
Specific risks: trend systems lose in ranges, and leverage multiplies both the move and the fees. At 2x, a 6% stop is about 12% of that market's margin. The model does not simulate liquidation: a position's mark is floored at 5% of its margin.
Funding-rate carry
What it trades: delta-neutral pairs: spot long plus an equal perp short on Hyperliquid, collecting the funding longs pay shorts.
How it decides:
- 1On each re-rank, keep markets trading at least the min daily volume, shortlist the top three times "markets at once" by current funding, plus anything held.
- 2Rank them by average funding over the funding-average window, as an APR, and keep those at or above the min funding APR.
- 3Hold the top N at equal weight. A held market stays unless a replacement pays at least the min APR gain more.
- 4Half of each position is the spot long and half the perp margin, so funding is earned on half the capital. Each move costs 0.0725% of the amount traded plus gas.
| Setting | Default | Range | Meaning |
|---|---|---|---|
| Min funding APR | 8% | 2% to 50% | Only carry markets paying at least this |
| Markets at once | 3 | 1 to 5 | How many markets to spread across |
| Min daily volume | $100M | $10M to $1,000M | Only markets this liquid |
| Re-rank every | 8 hours | 1 to 48 hours | How often to look for better funding |
| Funding average | 168 hours | 1 to 168 hours | Rank by this average, not the last hour |
| Min APR gain to switch | 15% | 0% to 30% | Keep a market unless a replacement pays this much more |
Specific risks: funding can turn negative, and then the position pays. The model treats spot and perp as offsetting exactly, so basis moves and a squeeze on the short leg are not captured. Rotating often eats the carry in fees.
Blue-chip basket
What it trades: a fixed basket: BTC 35%, ETH 30%, SOL 10%, LINK 10%, AAVE 10%, UNI 5%.
How it decides: it rebalances to target weights on schedule, or early when any weight drifts from its target by more than the drift trigger (measured relative to the target). With the trend filter on, it sells the whole basket to stables when BTC trades more than 1% below its average over the window, and does not buy back in until BTC is more than 1% above it.
| Setting | Default | Range | Meaning |
|---|---|---|---|
| Rebalance every | 30 days | 7 to 90 days | Scheduled rebalance |
| Drift trigger | 20% | 5% to 50% | Rebalance early past this relative drift |
| Trend filter | 0 days (off) | 0 to 120 days | Hold only while BTC is above its average |
Specific risks: with the trend filter off, this is a long-only crypto portfolio with full market drawdowns.
Retired templates and meme trades
Two token-filter templates, Meme momentum and Launch sniper, are no longer offered on /create. Agents created with them keep running on their rules. An engine agent's rules can also carry "meme trades on the side": one token at a time using the old Meme momentum filter, 2.5% of the vault a trade by default (1% to 50%), optionally from a set-aside share of 5% to 90%. This is not offered on /create; it only arrives through an approved rule change.
These templates run on testnet with paper positions. Results reported to testnet vaults are not real-money returns, and mainnet capital is planned, not live.
Related
Neuron is on Robinhood Chain testnet. Testnet performance is not evidence of real-money profitability.