Transform your trading with intelligent, algorithmic AI

No. The IQ Wizard exposes the whole pipeline — data, features, targets, signals, ML, backtest — as visual building blocks, and IQ Chat takes the same instructions in plain language. There is no point in the path from market data to a live bot where you have to write code. If you do write code, you can bring your own rules and models in instead.

The Free tier gives you 20,000 compute Tokens a month — enough for hundreds of small backtests — the IQ Wizard fully metered, 2 GB of storage, and full access to the Bot Advisor and marketplace listings. You can research and paper trade indefinitely. Running a live bot on your own exchange account starts at the Pro tier.

One Token is one CPU-second on our reference node, for jobs using up to 4 GB of memory. A GPU second costs 40 Tokens, and a job above 4 GB pays 1 + (GB − 4) / 8 per second. Nothing is fractional. A two-pair, one-year, hourly backtest costs single-digit Tokens; a heavy training sweep costs orders of magnitude more. Every job shows its estimate before it runs and its actual charge after, computed with the same arithmetic.

Your monthly allowance is a lot that expires with the period, so it is spent first. On top of it you can buy compute packs, which never expire and are only drawn on once the allowance is gone — paid credit never drains while included credit sits unused. You can also switch on pay-as-you-go with a trigger, a reload amount and a monthly ceiling you set yourself; a declined card stops the reloads rather than retrying. Pay-as-you-go is not available on Free.

So your job starts when you press the button. A monthly allowance with no burst ceiling can be spent in an afternoon, which is a capacity problem rather than a billing one. A quarter of your month is spendable in any rolling week, and a quarter of that in any rolling five hours. Both windows slide continuously rather than resetting on the clock.

No. You connect an exchange API key created without withdrawal or deposit permission, and your funds stay in your own exchange or broker account throughout. Orders cross the boundary; funds never do. QuantiX is a non-custodial software provider — we take no discretionary control over your account and cannot move your balance.

Two parts. A flat access fee of about $10 per deployed bot per month, and a 20% performance fee charged only on high-water-mark profits made on the capital you allocated to that bot — split equally between the strategy creator and the platform. If the bot makes no new profit, there is no performance fee. Fees are debited each billing cycle; an unsettled fee detaches the strategy rather than accruing a debt.

You are charged on new profit only. If a bot gains, you pay the performance fee on that gain; if it then loses and recovers, you pay nothing on the recovery — only on profit above the previous peak. You never pay twice for the same gain.

An upgrade takes effect immediately and is prorated, with the billing date reset. A downgrade or cancellation takes effect at the end of the period you have already paid for, and you can reverse it any time before then. There are no lock-in mechanics. You can pay by card or in crypto: card renewals are automatic, crypto renewals are a prepaid invoice per period.

Anything that describes a candle using that candle and the ones before it — open, high, low, close, volume, and the indicators built from them. Features are numeric, like RSI or MACD, or categorical, like a crossover label. A target is not a feature, because a target looks at candles that come afterwards.

A target is a function of future candles that scores how an asset actually performed — a number, or a label such as Buy, Sell or No-Action. Because it reads candles that had not happened yet at decision time, you cannot trade it directly; it exists to train models and to measure whether a feature carries any signal. QuantiX ships 11 target families, including Triple Barrier, fee-adjusted stop/take classification and ATR-normalised barriers.

A set of conditions combined with AND and OR that triggers entries and exits in a backtest and then in a live bot. You can compose signals from feature labels by hand, threshold a trained model's output, or mix the two — globally or per pair.

Two numeric gates, and neither is advisory. Before a backtest runs, the agent checks that ROC-AUC and PR-AUC train/test gaps stay within 0.25, that signal precision beats random, and that gain ratio is at least 1.0 — a failure rewinds the run to the node that caused it. After the backtest it checks expected return above 1%, Sharpe above 1, worst expected drawdown better than −20% and recovery under 150 days. Look-ahead leaks, including training and backtest overlap or a target used as a feature, are rejected before the run starts. Every live bot then renders the same 24 reports its backtest produced, so you can read live against backtest on identical terms.

Yes to both. Existing rules and models drop into the same pipeline and are judged by the same backtester, and you can upload your own CSV data. IQ Chat is included on Pro, and Max gives five times Pro's IQ Chat usage, on a meter kept separate from compute. You can instead connect your own key across six providers for a small handling fee.

Eleven crypto venues. Binance, Bybit, OKX, KuCoin and Gate.io run both spot and futures; HTX, BingX, BitMart, MEXC, CoinEx and XT run spot. Connecting one takes minutes — an API key and secret with trading permission and no withdrawal rights — and connecting accounts is unlimited and free. Kraken is scheduled for Q1 2027.

Research and backtesting are live today across forex history back to 2008, 14,000 active and 7,000 delisted US equity tickers, 130 futures and 115 indices, using the same engine, the same validation gates and the same reports as crypto. Live forex execution runs through the cTrader Open API: QuantiX is a registered cTrader application, so one integration reaches every cTrader broker rather than tying you to one. That execution path is in active development for Q4 2026.

Each live bot is its own isolated container with its own outbound IP, so another customer's load or rate limits never reach your account. Every entry, exit and stop fires a real-time Telegram alert, and you can monitor positions and stop a bot from either the platform or Telegram.