Beginner-friendly bots

The ladder, not the leaderboard

Bot	Type	Price	Free tier	Trial	Backtest	Paper	Trade fee	US
Bitsgap	Cloud platform	$23-149/mo	yes	7 days, no card	yes	yes	-	US
Coinrule	Cloud platform	0.02% of volume	yes	-	-	-	-	US
Cryptohopper	Cloud platform	$29-129/mo	yes	3 days, no card	yes	yes	-	US
PI Pionex	Exchange-native	0.05% per trade	yes	-	-	-	0.05% per trade	US
WunderTrading	Cloud platform	$0-100/mo	yes	7 days, no card	-	yes	-	US
3Commas	Cloud platform	$15-140/mo	-	3 days, no card	yes	yes	-	US
MI Mizar	Cloud platform	0.1% of volume	yes	-	yes	-	-	-
Binance Trading Bots	Exchange-native	free	yes	-	-	-	-	-
OKX Trading Bots	Exchange-native	free	yes	-	-	-	-	-
GA Gate.io Trading Bots	Exchange-native	free	yes	-	-	-	-	-
KuCoin Trading Bots	Exchange-native	free	yes	-	-	-	-	-
Bitget Trading Bots	Exchange-native	free	yes	-	-	-	-	-
BY Bybit Trading Bots	Exchange-native	free	yes	-	-	-	-	-
TR TradeSanta	Cloud platform	$14-70/mo	yes	14 days, no card	-	-	-	US

Live values from our database - details and caveats in each review. "-" = not documented.

Beginner advice usually reads like a leaderboard. It should read like a ladder: each rung exists to make a specific class of early mistake cheap.

• Paper first. Every platform on this page simulates - WunderTrading and 3Commas have paper modes, Freqtrade's dry-run is the gold standard[3]. A strategy that cannot survive simulated drawdown does not deserve keys.
• Exchange-native second. Grid and DCA bots inside Pionex (0.05%[1]), Binance, OKX or Gate are free, need no API keys, and execute on the venue's own matching engine. The fee you pay per trade[4] is the whole cost of tuition.
• Cloud trial third. Only when you can name the missing feature - cross-exchange logic, TradingView webhooks, marketplace access - take a trial (WunderTrading's 7-day[2], 3Commas' 3-day) and test the limits, not the dashboard.
• Self-hosted last. When ops habits exist, the self-hosted lane hands you the ceiling of control.

The two losses beginners actually take

Our drawdown statistics on real BTC data (2017-2026) frame both[7]:

Safety-order multipliers without capital math. The aggressive 2.0x DCA ladder needs $25,500 to cover the drawdown a flat 1.0x ladder covers with $800 - a 32x difference beginners discover mid-crash. The DCA hub carries the full capital-lockup table; the one-line version: compute the ladder's total requirement before enabling multipliers, not after the second fill.

Oversizing "because the backtest looked safe". BTC printed 22 drawdowns of -20% or worse since 2017, worst -76.6%[7]. Any backtest that does not include a regime like that is advertising. The bot-vs-hodl guide shows the same strategy scoring wildly differently across five regimes - the beginner-friendly conclusion is humility about parameter portability, and small size as the antidote.

The scam filter, step by step

Beginners are the target audience of every funnel in this space, so the filter deserves its own rungs. First: any platform that found you through an ad, a DM or a "mentor" is pre-flagged - legitimate software does not recruit in Instagram comments, and the FCA's bot-branded warning entries document exactly these channels[6]. Second: any pitch that names a return is finished - the FTC's line is that only scammers promise fixed profits[5], and "AI" prefixes change nothing (our AI lane has the full taxonomy). Third: any "manager" who offers to run the bot for you is a counterparty holding funds - the exact structure the FCA warns about[6], and the one where recovery scams double the damage.

The positive filter is shorter: the platforms on this page publish fees, limits and company details that our fact-checks verify against live pages, and their communities document failures as often as wins. Transparency asymmetry is the beginner's fastest tell - the scam-check tool screens a site in seconds and costs nothing.

One more calibration helps: compare every promise against what the honest community tests achieved. The best-documented free-bot experiment in our corpus - 111 days on Pionex native bots during the 2022 bear - returned +1.1% while BTC itself fell -16.6%[8]. That is a real, documented, fee-inclusive result from free tooling in a hostile market: modest, regime-dependent, and proud of neither being more nor less. Any pitch promising multiples of that without a regime table is pricing your inexperience; any result smaller than that without fees disclosed is hiding the stack. The bot-vs-hodl guide gives you the regime framework to judge both. The same corpus carries the failures - bags accumulated by ladders that ran out, fees that turned wins into losses - and they are the calibration data a beginner cannot buy anywhere else.

The scam layer, because beginners are the target

14 platforms pass this page's beginner filter, and the pool is pre-filtered: 13 are free or freemium, so no upfront payment is ever needed to start. Everything that requires an upfront payment - "pro bot licenses", "AI signal packages", pools promising fixed returns - sits outside this comparison and inside the FTC's fraud patterns[5]: the fixed-profit line is the oldest one on the internet. The FCA's warning list shows bot-branded unauthorised firms recruiting through Instagram and Telegram[6]; our scam patterns guide walks the full funnel.

Two beginner-specific rules complete the filter: never share API keys with withdrawal permissions (the key security guide is the playbook), and treat any "manager" offering to run a bot for you as a counterparty holding funds - the exact structure the FCA warns about[6].

A first week, scripted

The concrete version of the ladder, for the reader who wants instructions: day one, pick the venue you already use and find its native bot section - set up a paper or minimum-size grid on BTC or ETH against a stablecoin, wide spacing, no leverage, no futures toggle. Day two, write down what the bot is supposed to do (entries per day, expected inventory at range bottom, what you will do at breakout) - a strategy you cannot state in five sentences is not configured yet. Days three to seven, check twice daily, make no parameter changes, and watch fills rather than PnL. The point of week one is not profit; it is calibration of your own attention.

End the week with two numbers: total fees paid and largest unrealized drawdown. Fees calibrate the cost awareness that every later decision needs[4]; the drawdown calibrates size - if it felt large, it was, and the next allocation should be smaller, not the same. That loop (state intent, run untouched, measure fees and drawdown, adjust size) is the entire beginner curriculum, and it transfers unchanged to every platform on this site.

Warnings

• Automation does not remove drawdowns - it removes panic. The same -53% BTC episode that is open through 2025-26 hit every bot and every hold alike[7]. Size for it; do not automate around it.
• Freemium limits distort risk. One free bot tempts overloading it with aggressive settings; the constraint should limit size, not inflate per-bot risk[2].
• Fee awareness is the first skill. 0.1% per fill[4] on frequent strategies compounds into a real hurdle - the grid and DCA hubs quantify it per strategy.
• Community test numbers are honest, not typical. The 111-day free-bot test (+1.1% while BTC fell -16.6%[8]) is documented with methodology; screenshots in ads are not. Ask for the drawdown, the window, and the fee assumptions of anything promising more.
• Trends beat simple bots. Our regime backtests show holding outperforming simple automations in bull runs[7] - the beginner's honest expectation for a first grid is "smoother than holding", not "more than holding".

What the numbers say

• 14 platforms pass the beginner filter; 13 of them are free or carry a free tier, and 5 offer a trial - the first month should never cost money.
• Trial windows run up to 14 days - long enough to paper one full drawdown if you actually use the window.
• 50% document US availability - jurisdiction filters beginners the same as experts.
• The how bots work guide explains the execution loop every option shares, so the first subscription - if it ever comes - is bought for a named feature, not a name.