What HyperDCA actually does

HyperDCA places orders on an exchange account that stays yours. You give it trading credentials; it never gains the ability to withdraw, and your balance never leaves the venue. There are two engines and they can run side by side:

• DCA. Scheduled recurring buys. You set the instrument, the amount and the interval. Sizing can lean heavier as price falls below your base, and each buy is stress-tested against a worst-case reversal before it is placed.
• Autopilot. An AI model reviews the market on your schedule and proposes actions — open, close, take partial profit, move a stop. Every proposal is validated against your risk limits in code before anything reaches the exchange.

Before you start

• An account on one of the ten supported exchanges, funded, with derivatives enabled.
• The ability to create an API key or agent key on that exchange (see the table below).
• Optionally, your own API key for an AI provider. Autopilot supports 20 models across 9 providers — Anthropic, OpenAI, Google, xAI, DeepSeek, Moonshot, DeepInfra, Groq and OpenRouter. You can bring your own key rather than using a shared one.
• Capital you can afford to lose entirely. Start smaller than feels interesting; you can raise the allocation once you have watched a few cycles.

1. Create your account

Sign up with an email address and password from the Get started page. If you arrived through someone's referral link the code is captured automatically and shown on the form before you submit.

Immediately after signing up you are taken through a short setup wizard covering the next three sections. Every step in it can be skipped and finished later from Settings — skipping does not leave anything half-configured, it just means the engine has nothing to run yet.

2. Connect an exchange

Ten exchanges are supported. Which credentials you need depends on the venue, and several of them have a specific gotcha worth knowing before you go looking for the settings page.

When you connect, HyperDCA immediately calls the venue to confirm the credentials work and reports the balance and market count it can see. If that number is not what you expect, stop and fix it before going further — it is what every sizing decision is based on.

You can also set a fixed allocation cap in dollars at connect time. Leave it blank and the engine tracks your full account equity; set it and the engine behaves as if that is all the capital it has. Setting one is the simplest way to limit your exposure on your first run.

3. Choose a strategy

A strategy is a set of written instructions telling the AI what kind of trades to look for: how aggressive to be, what to avoid, when to take profit. There are 39 built-in strategies covering accumulation, momentum, mean reversion, and options approaches like the wheel and covered calls.

You can fork any of them, edit the instructions and run your own version. Whatever is active is shown on the dashboard, and each decision the engine logs records which strategy produced it.

4. Set your risk limits

These are the numbers that matter most. Each has a conservative default and an absolute ceiling that cannot be exceeded no matter what you or a strategy asks for.

Alongside those, a few rules are always on and are not configurable:

• Every opening trade must carry both a take-profit and a stop-loss, or it is rejected.
• Breaching the drawdown limit force-closes every position. Closes, partial closes and stop moves keep working during a breach — the emergency exit must never be blocked by the same limits that caused it.
• A confidence floor, a daily trade cap and a cooldown after a loss all apply to opening trades, so a bad run cannot turn into a fast run.

You can also cap positions and exposure in absolute dollars in addition to the percentages. Where both apply, the smaller one wins.

5. Turn it on

Once an exchange is connected and a strategy is active, choose how often the engine reviews:

• Fixed. A review every N hours, exactly. The most predictable, and the cheapest to run.
• Dynamic. The interval moves within a floor and ceiling you set, based on what the market is doing.
• Aggressive. Reacts to price movement between scheduled reviews. Uses more AI credits.

You do not have to wait for the first scheduled review to see the engine work. Force Review on the dashboard runs a full cycle immediately and writes the result to the decision log, so you can read exactly what your strategy proposed, what the guardrails allowed, and what it cost — before the schedule ever fires on its own.

What happens on each review

1. A fast, cheap model screens the instrument universe and shortlists candidates.
2. If enabled, a web search pulls recent news and the current macro regime for those names.
3. Your chosen model does the deep review with your strategy's instructions and your open positions in context, and returns proposed actions.
4. An allocator resizes anything too large to fit your budget rather than discarding it.
5. Guardrails validate every action. Anything that fails is blocked, and the reason is recorded on the dashboard.
6. Whatever survives is placed on the exchange, and the whole cycle is written to the decision log with the model used and the cost.

Every one of those steps is visible after the fact. The activity feed on the dashboard shows what was proposed, what was blocked and why — start there when the engine does something you did not expect.

How your keys are handled

• Credentials are encrypted before they are written to storage and are never returned by the API once saved. You cannot read them back, and neither can support.
• On Hyperliquid, use an agent key. It is scoped to trading and cannot withdraw, so a compromise of our servers still cannot move your funds.
• On the centralised exchanges, create the key with the narrowest permission set that still allows trading, and never enable withdrawal. Restrict it to an IP where the exchange supports that.
• Use a key that is not shared with anything else. If you need to revoke it, revoking on the exchange takes effect immediately and does not need us.
• Sessions use short-lived tokens that renew silently. Signing out everywhere revokes every device at once from your profile page.

Troubleshooting

The balance shows $0.00

Usually the credentials are valid but pointed at the wrong sub-account, or the funds sit on a spot wallet rather than the derivatives wallet. On Hyperliquid, check the funds are on the account the agent key was generated for. On KuCoin and Kraken, check you created the key on the futures domain rather than the spot one.

Nothing has traded

Check, in order: auto-trading is on; a strategy is active; the review interval has elapsed; and the activity feed for blocked actions. A block with a stated reason is the engine working correctly — it is telling you which limit is binding.

Every action is blocked for confidence

Your minimum-confidence floor is above what the model is producing for current conditions. Lower the floor or accept fewer trades. Do not fix this by removing the stop-loss requirement; you cannot, and you would not want to.

Connecting fails with an authentication error

Most often a missing passphrase (OKX, Bitget, KuCoin), a spot key used where a futures key is required (KuCoin, Kraken), or the futures permission not enabled on the key (Binance, Gate.io).

Risk

Trading perpetual futures and options with leverage carries a high risk of loss and is not suitable for everyone. You can lose your entire deposit. The drawdown kill switch and the other limits described here reduce how badly a single run can go; they do not make the activity safe, and they cannot protect against a venue outage, a gap through your stop, or a strategy that is simply wrong about the market.

HyperDCA is a software tool. It is not a broker, not a custodian, and does not provide investment advice. Nothing on this page is a recommendation to buy or sell anything, and you remain responsible for everything your exchange account does.