TL;DR: There is no honest "best Hyperliquid trading bot" ranking, because the right tool depends on your strategy and how much control you want. What you can compare objectively is how each bot handles custody and key permissions, how honest its backtests are, what it really costs per trade, and how it behaves when something breaks. Use the checklist below on any bot, including Trigr, before you approve a key.
Why a checklist instead of a ranking?
Search for the best Hyperliquid bot and you will find lists that rank products by features or claimed returns. Rankings based on returns are especially weak: past performance depends on the period, the leverage and the strategies chosen, and a screenshot of a profitable month says nothing about the next one.
The things that decide whether a bot is safe to use are more boring and easier to check. Can it withdraw your money? What exactly did you authorize? Do its backtests include costs? What happens when an order fails? A bot that answers these well might still lose money, since markets are uncertain, but it will not lose money for reasons you did not see coming.
1. Custody: who can move your funds?
This is the first question, and it has a clear right answer on Hyperliquid.
Hyperliquid supports API wallets, also called agent wallets. According to Hyperliquid's documentation on API wallets, a master account can approve API wallets to sign on behalf of the master account or its sub-accounts. Withdrawals are signed by the master wallet, not by an API wallet. A well-designed bot asks you to approve an API wallet with one signature and never touches your main key.
What to check:
- Does it ask for your seed phrase or main private key? If yes, stop. Anyone holding that key can withdraw.
- Does it use an API wallet? Confirm the approval in your wallet shows an agent approval, not something else.
- Who holds the API wallet key? Some tools keep it in your browser; others store it server-side so they can trade while you are offline. Both are valid designs, but you should know which one you are choosing.
- Can you revoke it? You should be able to revoke or replace the approval yourself.
The article on Hyperliquid agent wallets goes deeper on how the approval works.
2. Key permissions and expiry
API wallets are not permanent by default. Hyperliquid's docs note that wallets can expire or be pruned, and that registering a new unnamed API wallet deregisters the previous one. In practice, approving a second bot, or the same bot from another device, can silently break the first.
Ask each bot:
- How long does the approval last, and does it warn you before expiry?
- Does it detect when its key has been replaced, and stop cleanly rather than failing order by order?
- Can the key do anything beyond trading? Some platforms use the approved key for limited account actions such as moving USDC between your own main account and sub-accounts; ask what is allowed and what is not.
3. Backtest honesty
If a bot shows a backtest, find out how it was produced. The same strategy can look excellent or mediocre depending on a few assumptions:
- Fill timing. A signal computed on a bar's close should fill at the next bar's open, not at the close it was computed from.
- Costs. Trading fees at minimum. On perps, funding and slippage too, or at least a clear label that the result is gross.
- Point-in-time data. Indicators and outside data should only use what was known at the time.
- Intrabar ties. When a take-profit and a stop-loss are both hit in one candle, a conservative engine assumes the worse outcome or checks finer data.
- Overfitting. A strategy selected from many variants needs statistics that account for the search, such as the Deflated Sharpe Ratio.
- Forward record. Is live or paper performance shown separately from the fitted backtest?
The guide on how to read a backtest explains what to look for in each of these.
4. What does it really cost?
Costs on Hyperliquid bots come from several layers, and they compound with trade frequency.
- Exchange fees. Hyperliquid charges maker and taker fees per its published fee schedule.
- Builder fees. Hyperliquid's builder codes let the app sending your order charge a fee on each fill, capped at 0.1% for perps and 1% for spot. You approve a maximum per builder with your main wallet and can revoke it at any time.
- Subscriptions. Monthly plans, credit systems or performance fees.
- Slippage and funding. Not charged by the bot, but real, and they depend on how the bot trades.
Add these up at your expected trade count. A strategy that trades several times a day can lose most of its gross edge to fees alone.
5. Behavior when something goes wrong
The most revealing questions are about failure:
- Does every entry get a protective stop placed at the exchange, or does the stop live only in the bot's memory?
- If placing a stop fails, does the bot retry and flag it?
- Does it reconcile its records with the exchange regularly, so a position closed by the exchange is recorded correctly?
- What does it do if you trade manually on the same account?
- What does it do with orders below Hyperliquid's minimum size?
- Can you pause it, and what exactly does pausing do to open positions?
- Which alerts does it send, and through which channels?
The checklist in one table
| Area | Good answer | Red flag |
|---|---|---|
| Custody | API wallet approval; main key never shared | Asks for seed phrase or main private key |
| Key lifecycle | Known expiry, detects replacement, revocable | No expiry info; fails silently |
| Backtest fills | Next-bar open, stated clearly | Fills at the signal bar's close, or unstated |
| Backtest costs | Fees always; slippage and funding available and labeled | Gross results presented as real returns |
| Overfitting | Trial-aware statistics, separate forward record | Only a best-case curve |
| Fees | Builder fee and plan costs published | Costs discovered after the first trades |
| Failure handling | Exchange-side stops, reconciliation, alerts | Stops held only in memory |
| Account safety | Handles manual activity or requires isolation | Silently trades around unknown positions |
How does Trigr answer the checklist?
For transparency, here is how Trigr answers each point. Use the same scrutiny you would apply to any other bot.
- Custody. A Hyperliquid agent uses a trade-only agent key (an API wallet) that can place and cancel orders and cannot withdraw or send funds to another address. For autonomous agents, Trigr stores that trade-only key encrypted server-side; you keep the master wallet key and custody of funds. The Connect Hyperliquid docs describe the flow.
- Key lifecycle. The autonomous key is valid for about 90 days. If it has been overwritten from another device, the agent pauses in a "needs reconnect" state.
- Backtests. Point-in-time data, next-bar-open fills, and the stop wins an intrabar tie that 5-minute replay cannot resolve. Trading and builder fees are always applied; slippage (a flat bps per fill) and funding (historical or flat per-8h) are opt-in, and a first result is labeled gross. DSR and PBO come from ML optimization runs; a plain backtest labels them as not yet computed.
- Costs. The Hyperliquid builder fee is up to 0.05% per trade. The fees docs explain what is charged per trade; plans run from Free to Trader ($29 per month) and Pro ($99 per month), and live trading is available on every plan.
- Failure handling. Every live entry ships with reduce-only take-profit and stop-loss orders; failed protection is retried and flagged; agents reconcile against Hyperliquid every minute; alerts go by email or web push.
- Account safety. If Trigr detects activity it did not place on the account, it safety-pauses the bound agents and sends a mandatory email. Hyperliquid rejects orders under $10 notional, so entries below $11 are raised to $11.
- Forward record. Each agent's chart has a frontier marker between backtest and forward record, and marketplace strategies are immutable, versioned, with verified backtests.
None of this makes trading safe in the sense of profitable. Backtests are not guarantees; perps are leveraged and can lose more than expected.
When might another bot suit you better?
Trigr is one approach, and not the right one for every trader. A self-hosted open-source bot may fit better if you want to write code and run it on your own machine; the Freqtrade and Hummingbot comparison covers that route. A market-making bot fits better if you want to provide liquidity rather than trade signals. And if you mostly chart and test ideas across many markets, the TradingView comparison is a better starting point.
Next steps
Run the checklist on any bot you are considering. If you try Trigr, the pillar on deploying a strategy as a Hyperliquid agent walks through the path from backtest to paper to live, and the strategy marketplace lets you inspect verified backtests before you connect anything.