Trading Bot Checklist: 25 Checks Before You Go Live

TL;DR: Most live-bot failures are not bad ideas; they are unchecked assumptions about costs, sizing, keys and what happens when something goes wrong. Before a trading agent touches real money, verify five things in order: the strategy survives realistic costs, a paper forward test behaved like the backtest, the key can trade but not withdraw, every order clears the venue minimum at your sizing, and you have alerts plus a written plan for pausing. The checklist below works for any bot; the notes explain how each item maps to Trigr.

Why use a checklist at all?

Pilots use checklists not because they forget how to fly, but because routine steps get skipped under time pressure. Launching a bot is similar. The strategy is the interesting part, so attention goes there, while the boring parts (key permissions, minimum order sizes, what "pause" actually does) cause the expensive surprises.

A checklist also forces you to write down what you expect. If you know the backtest averaged three trades a week, a week with zero trades is information rather than panic.

The items are grouped by phase. Tick every box before you go live, and treat any box you cannot tick as a reason to wait.

Phase 1: Is the strategy worth running?

1. The backtest is positive net of costs. Trading fees are always applied in a Trigr backtest, but slippage and funding are opt-in and off by default, so a first result is gross. Rerun with a slippage assumption you would defend and historical funding switched on. See slippage and funding in perp backtests.
2. The backtest is point-in-time. Signals use only data available at the bar's close and fill at the next bar's open. If you built the strategy elsewhere, check that it does not read the current bar's close and fill at that same price.
3. There are enough trades to judge. Thirty trades over five years says very little. Look at the trade log, not only the equity curve.
4. You know how many variants you tried. If this is the best of dozens of versions, expect it to underperform live. An ML optimization run reports the Deflated Sharpe Ratio, which adjusts for the number of trials; a plain backtest does not compute it.
5. You did not treat the recent period as a holdout. Studio shades the trailing 25% of a backtest as a recent-period diagnostic, but it is not out-of-sample data; standard backtests use full available history.
6. Every strategy has a defined exit. Trigr agents reject a strategy with no way out of a position: if you turn off the exit on an opposite signal, you need a trailing stop, a time stop, or both a take-profit and a stop-loss.

Phase 2: Did the forward test behave?

1. A paper agent has run the exact configuration. Same strategies, same allocations, same risk settings. A paper agent fills at the current Hyperliquid price and deducts the base taker fee and builder fee, with no slippage or funding.
2. Signal frequency matches the backtest. Roughly the same number of trades per week, on the bars you expected.
3. Direction and exits look right. Longs where you expect longs, stops and take-profits closing trades at plausible levels.
4. You compared against the right baseline. Compare paper results to a backtest run with costs, and remember paper still omits slippage and funding. A few weeks of an intraday strategy, or a few months of a swing strategy, is a reasonable minimum.

On a Trigr agent's chart, the frontier marker separates the backtested segment from the forward record. Everything to the right of it happened after you pressed start.

Phase 3: Are the account and keys set up safely?

1. The key cannot withdraw. On Hyperliquid, an API wallet (agent key) can place and cancel orders, while withdrawals require the master wallet's signature, as Hyperliquid's API wallet documentation describes. For autonomous agents, Trigr stores a trade-only key encrypted server-side; it cannot withdraw or send funds to another address. You keep the master wallet and custody. The agent wallet explainer covers the details.
2. You know how to revoke it. Know where to revoke the agent key from your master wallet before you need to.
3. You know when it expires. Autonomous agent keys are valid for about 90 days. An expired or replaced key puts the agent into a "needs reconnect" state.
4. The account is exclusive to the agent. Assigning an agent reserves that account for Trigr execution, even while paused or flat. If Trigr detects a manual trade or another bot on it, it safety-pauses every agent bound to that account and emails you. Do not trade the same account by hand.
5. You considered a subaccount. Hyperliquid only allows subaccounts after $100,000 of trading volume, per its sub-accounts documentation. Once eligible, a dedicated subaccount gives stronger isolation. The agent's capital figure is a sizing limit, not a wallet boundary.
6. The account is funded. Hyperliquid will not credit deposits below 5 USDC. Adding collateral while the agent runs is fine; do not withdraw or transfer it out while managed positions are open.

Phase 4: Does the sizing work at the venue?

1. Every entry clears the minimum order size. Hyperliquid rejects orders below $10 of notional, so Trigr raises smaller agent entries to $11. That makes a tiny sizing rule take a larger position than its percentage suggests, and an agent with under $11 of capital opens nothing. Work out allocation times size times leverage for each slot.
2. Peak exposure is acceptable. Max concurrent positions per strategy and across the agent set the worst case. Trigr's setup shows a risk panel with total exposure and per-asset exposure, and flags over-leverage.
3. A bad day is survivable. Estimate what happens if every open position hits its stop at once. Correlated crypto perps often move together.
4. The market is live-eligible. TradFi markets listed through Hyperliquid HIP-3 (gold, the S&P 500, single stocks) can be backtested and run on paper agents, but live execution on them is not available in the current beta.

For how size, leverage, ATR trails and time stops interact, see the position sizing guide.

Phase 5: Will you know when something goes wrong?

1. Alerts reach you. Trigr sends alerts by email and web push for signals, take-profits and stop-losses, and agent events such as opened or closed positions, errors and liquidations. Turn on the ones you need in settings and test that push works on your device; the alerts docs explain the iOS install step.
2. Safety emails are not filtered. An account-safety incident always sends its email even if you muted general agent-error emails. Make sure it does not land in spam.
3. You have a review schedule. Decide in advance when you will look: daily for the first week, then weekly. Write down which numbers you will compare. Monitoring a trading agent goes deeper.

Phase 6: Do you have an exit plan?

1. You know what pause does. Pausing a Trigr agent blocks new entries, while positions it already manages can still close through their exits or reconcile. Pausing is not flattening. If you need to be flat, pause first, then close positions.
2. You wrote down your stop conditions. For example: pause if live drawdown exceeds the backtest's worst drawdown, if trade frequency differs sharply from the backtest for a month, or if fills are consistently worse than your slippage assumption. Decide these before you are emotionally involved.

The checklist at a glance

What changes for a Propr prop-firm account?

The same list applies, with two differences. The key is a Propr API key that can place and close trades but not withdraw, because Propr's API offers no withdrawals and Propr custodies the funds. And the challenge's daily-loss and drawdown limits become the binding constraint, so Trigr replays your backtest under that account's rules before deployment. Running a prop-firm challenge with an agent covers that flow.

Live trading on Trigr is available on every plan, including Free. Start small, scale up only after live fills, fees and P&L match your expectations, and remember that backtests are not guarantees and leveraged perps can lose more than you expect.

Next steps

Work through the full deployment path in from backtest to a live Hyperliquid agent, or browse verified strategies on the marketplace to paper-test before you build your own.