How to copy trade on Hyperliquid in 2026: a complete guide

What copy trading on Hyperliquid actually means

Copy trading is the practice of reproducing another account's positions in your own. On Hyperliquid this is unusually tractable, because the exchange publishes fills, positions and balances on-chain. You can see what a leader did, when, at what size and at what price, without asking them and without trusting a reporting layer.

What is public is the outcome, not the intent. You see that an account opened 12 ETH long at a given price. You do not see the stop they had in mind, the correlated position they may hold elsewhere, or whether this is a directional bet or one leg of a spread. Copy trading always reproduces the observable half of a decision.

There is also a timing gap that no system removes. The leader's order fills first, your mirror order fills afterwards, and the difference between those two prices is yours. On a liquid major during normal conditions, that difference is small. During a volatility expansion in a thin market, it is not.

Everything below assumes you already understand that perpetual futures are leveraged instruments. Copying someone else's leverage does not make it less leveraged, and liquidation risk transfers along with the position.

The three main approaches

Almost every copy trading option on Hyperliquid reduces to one of three structures. They are genuinely different in what happens to your money, so it is worth being precise rather than treating them as variants of the same thing.

Native vaults: pooled capital, one strategy

Hyperliquid supports vaults, where a leader trades a pooled balance and depositors share the result in proportion to their deposit. Copying is exact by construction, because there is nothing to mirror — you are in the same book.

The trade-offs are structural. You deposit, which means your capital leaves your own account and sits in the vault. You get one strategy: whatever that vault runs, with no ability to weight it against others inside the same position book. Exit is subject to whatever lockup or withdrawal terms the vault defines. And a vault's leader is a single point of failure in exactly the way a single copied leader is.

What you gain is simplicity and zero tracking error. For someone who has done the work on a specific leader and wants exposure to precisely that leader, a vault is a coherent choice. It is not diversification, and it is not non-custodial in the sense of your funds staying in your own account.

Custodial or API-key tools: convenience with counterparty exposure

The second category is third-party software that trades on your behalf — either by holding funds directly, or by holding credentials that can act on your account. This is the most common structure in centralised-exchange copy trading and it carries over to Hyperliquid.

The question to ask is narrow and answerable: what exactly can this software do that I cannot undo? If it holds funds, you have counterparty risk in the ordinary sense. If it holds a credential, the relevant question is the scope of that credential — whether it can only trade, or whether it can also move capital.

This category is not inherently bad, and dismissing it wholesale would be dishonest. It is a spectrum, and the position on that spectrum is determined by permission scope and by how easily you can revoke. Ask for both in writing before funding anything.

Non-custodial agent systems: trade-only permission

Hyperliquid has a native mechanism for this: agent approval. You sign an approval that authorises a specific agent address to place and manage orders on your account. The approval is a permission, not a transfer. Your USDC never leaves your own Hyperliquid account.

The scope is the point: trade-only agent approval; no withdrawal or transfer rights. Withdrawals and transfers remain exclusively signable by your own wallet, which means the software running the strategy cannot move your capital anywhere even if it wanted to, and even if it were compromised. You can revoke the approval on-chain at any time without asking anyone's permission or waiting for a support ticket.

Fees in this model are typically charged through Hyperliquid's native builder-fee mechanism, where you approve a maximum rate up front and it is applied to mirrored volume. That approval is also visible and capped before any trade is placed.

The remaining constraint is account structure. If a system mirrors several leaders into one account, their positions net against each other and you are no longer diversified regardless of custody model. Solving that requires one sub-account per leader, which is why isolation and diversification are the same engineering problem.

Step by step: starting with a non-custodial agent system

The sequence below is the one HyperMirror uses, but the shape is general — any agent-based system on Hyperliquid will look broadly like this, and the steps that are missing from a given product tell you something about it.

Step 1: fund your own Hyperliquid account, on the Perps balance

Deposit USDC to your own Hyperliquid account. Not to a product address, not to a vault — your account, under your address, visible on-chain.

The detail that trips most people up is the internal balance split. Funds sitting on Spot or in a unified balance cannot be used as perpetual futures margin. They have to be on the Perps balance before anything can trade. A readiness check that names this blocker explicitly is worth having; without one, the usual failure mode is an account that looks funded and mirrors nothing.

Step 2: approve a trade-only agent

Sign the agent approval. Before you sign, read the payload: it should name a specific agent address, and it should be an approval to trade, not a transfer authorisation.

If the agent has already been approved on a previous attempt, the system should detect that from your on-chain extra-agent list and skip the step rather than asking you to sign again. Being asked to re-sign something already granted is a small signal about how carefully the integration was built.

Step 3: approve a capped builder fee

Approve the maximum builder fee rate. HyperMirror's is 0.1% of mirrored notional volume, charged through Hyperliquid's native mechanism and applied to mirrored notional only. You approve the ceiling; it cannot be raised unilaterally afterwards.

Note what this fee is not: it is not a subscription, not a performance cut of your equity, and it grants no withdrawal rights. It also means cost scales with turnover, which has a real consequence — a high-frequency leader costs materially more to mirror than a low-turnover one at the same fee rate.

Step 4: start mirroring and watch the first rebalance

Once the agent is approved and the Perps balance is funded, allocation runs: capital is spread across the qualified basket in proportion to composite score, with each leader mirrored in one Hyperliquid sub-account per mirrored leader.

Below $100,000 of mirrored volume of mirrored volume, HyperMirror runs a single scored leader with the same controls; above it, up to 10 leaders run in parallel. The gate exists because a small account cannot open ten isolated sleeves and still clear minimum order sizes in each — sub-scale sleeves produce tracking error that swamps the benefit of diversifying.

The first hours are the ones worth watching. Check that each sleeve holds what you expect, that the position count matches the leader count, and that no sleeve is sitting unfunded.

Step 5: know how you stop

Before you start, know the exit. There should be at least two: a control inside the product that pauses mirroring, and an on-chain revocation of the agent approval that does not require the product to cooperate.

Revoking the agent stops new orders. It does not close existing positions — those remain yours to manage or close as you choose, which is the correct behaviour, since a forced liquidation of your book on revocation would be its own hazard. Plan for the position management, not just the switch.

The risks that do not disappear

No structure above removes the following. Anyone who tells you otherwise is selling something.

• Netting: multiple leaders in one account merge into one net position per market, destroying both the exposure you paid for and the per-leader attribution needed to score them.
• Leader decay: edges are regime-dependent and the PnL curve reports the change after it has been paid for. Rule-based replacement is reactive by design.
• Tracking error: your fills are not the leader's fills. Latency, size, book depth, funding and fees all separate your result from theirs, permanently and in both directions.
• Liquidation: isolation contains a liquidation to one sleeve; it does not prevent it. That sleeve's capital is still gone.
• Correlated deleveraging: in a market-wide stress event, ten leaders stop being ten independent bets and fall together.

How to evaluate any copy trading system

Ten questions. If a product cannot answer all of them plainly, that is itself the answer.

• Where do my funds sit while the system runs — my account, or somewhere else?
• What exactly can the permission I grant do, and what can it not do?
• How do I revoke, and does revocation require the product's cooperation?
• Do multiple leaders share one position book, or is each isolated?
• How are leaders selected, and can I check the criteria against live output?
• How is capital weighted across leaders, and why that scheme?
• What triggers a leader being removed, and how fast does it act?
• What is the total cost — fee, exchange fees, spread and turnover — not just the headline rate?
• What is published: per-leader attribution, or one blended number?
• What does the product say it cannot do? A system with no stated limits has undisclosed ones.

Common mistakes

Sizing to the upside. People choose capital based on what the record suggests they might make rather than what they can lose. In leveraged perpetuals the correct anchor is the loss.

Copying one leader and calling it diversification because the leader trades several markets. Market breadth inside one account is not the same as multiple independent processes.

Judging the system after a week. A fortnight of results is variance. If the selection method needs hundreds of closed trades to be meaningful, so does your evaluation of it.

Comparing your account to a published model figure without matching the window, the mode and the entry timing. Entry timing alone is usually the largest source of divergence.

Leaving funds on the wrong internal balance and concluding the system is broken when nothing trades.

Ignoring turnover. Two systems with the same fee rate can have very different total costs if one mirrors leaders who trade ten times as often.

Conclusion

The copying part of copy trading is the easy part. What separates these approaches is custody, account structure and how honestly the costs and limits are stated — and those are all checkable before you commit capital.

If the non-custodial route is the one you want to look at, the how-it-works page walks the same five steps in sequence, and the onboarding flow shows the exact permission text before anything is signed. Perpetual futures carry a substantial risk of loss, including the loss of your entire position.