Hyperliquid builder codes: who gets paid and how much in 2026

One hundred and two distinct interfaces—wallets, mobile apps, terminals, and bots—have captured $89,986,616 in fees from the same Hyperliquid order book without operating one themselves, according to live DefiLlama data from August 3, 2026. The mechanism is called builder codes: a parameter within the protocol itself that allows any interface to attach its address to every order and automatically collect an on-chain surcharge on the trade, without a matching engine, settlement, or proprietary market makers. In the perpetuals market (perpetuals: futures with no expiry date, the instrument used for leveraged trading on an asset's price), this turns the front-end into an autonomous business line. This article breaks down how much the top ten are actually billing, separates two mechanisms with different capital, risk, and distribution profiles—builder codes and HIP-3, the deployment of proprietary markets—, documents the revenue collapse of the front-end layer between July 2025 and July 2026, and settles a specific lingering doubt: the Based token was already released on March 30, 2026, and there is no pending airdrop associated with that issuance.

What is a builder code and how does an interface without an order book get paid?

Hyperliquid documentation clarifies from the outset that "builder" here does not refer to consensus block builders, but rather to those building applications on top of the chain. The mechanism fits into a two-field object that travels within each order: {"b": address, "f": number}. The b field is the builder's address and f is the fee they wish to charge, expressed in tenths of a basis point—a value of 10 equals one basis point, or 0.01%. The fee is processed entirely on-chain as part of the fee logic and is claimed through the same circuit as referral rewards.

Before this works, the circuit requires three conditions:

1. The user approves the builder via the ApproveBuilderFee action, signed with their main wallet—not an agent or API wallet—with a maximum of ten active approvals at a time, revocable at any moment.
2. The builder maintains at least 100 USDC in value in their perpetuals account, using the standard account abstraction mode.
3. The protocol applies a cap: 0.10% on perpetuals—on both sides of the trade—and 1% on spot, where it only affects the sell side because builder codes are only collected on the quote or collateral asset.

Transparency goes further than usual: Hyperliquid publishes a daily compressed file with the detail of fills for each builder address at stats-data.hyperliquid.xyz/Mainnet/builder_fills/{address}/{YYYYMMDD}.csv.lz4. Anyone can audit which trades each interface routed and what they charged for them, without asking for permission or signing data agreements.

The structural consequence is what gives this analysis its title. The matching engine, order book, and settlement live in HyperCore, the component we analyzed in detail previously. Anyone wishing to bill on top of that engine does not need to replicate it: they simply need to get someone to sign orders from their screen. This is the "distribution over liquidity" thesis we proposed when analyzing Lighter's entry into Robinhood Wallet, taken one step further: there, two exchanges competed for the retail channel; here, a hundred applications that never built an exchange are competing.

How much is the Hyperliquid front-end layer actually billing?

The following table crosses, for the ten largest builders by cumulative fees, the volume of perpetuals they have routed, what they have charged for it, the implicit fee resulting from dividing one by the other, and their billing in the last thirty days compared to the previous thirty. All data comes from the same DefiLlama query on August 3, 2026, filtered by the Hyperliquid L1 chain and the interfaces category, using the aggregator's declared methodology: notional volume of all trades including leverage, and builder code fees on perpetuals trades.

A perimeter warning regarding the second row: the "Based" entry on DefiLlama groups several products from the same team—BasedApp, Based Alpha, Based Predict, and HYENA—while the perpetuals volume corresponds only to BasedApp. Taking BasedApp exclusively, cumulative fees are $15,095,022 and the implicit fee drops to 0.036%. The row uses the aggregated figure to maintain consistency with the rest of the table, which comes from the same query.

The aggregates provide the true measure of the distribution. DefiLlama records 103 interfaces with revenue on Hyperliquid; discounting trade.xyz, which operates under a different mechanism and deserves its own section, there are 102 builder code builders with $89,986,616 accumulated. The ten in the table concentrate 80% of that figure, and the top three account for 54.4%. The sector median stands at approximately $104,900 accumulated; fifty-one of the one hundred and two have billed less than $100,000 in their entire lifetime, and fifteen did not earn a single dollar in the last thirty days. The distribution is classic for any app market: a handful of channels with real audience and a very long tail of integrations that exist on the list but not in the order flow.

There is a comparison that puts the entire layer into perspective. In the last thirty days, the 102 interfaces totaled $4,062,622 in fees. In the same window, the Hyperliquid protocol itself captured $49,423,726. The distribution layer currently takes just over eight percent of what the engine it relies on earns.

How do builder codes differ from HIP-3?

The two mechanisms are often conflated under the label "Hyperliquid builders," yet they are distinct businesses in terms of capital, risk, and distribution. The following matrix separates them across their six operational axes.

The trade.xyz figure requires precision because it is the most frequently misquoted. DefiLlama counts $48,761,757 in cumulative fees in markets deployed by trade.xyz, with $8,806,199 in the last thirty days. That amount is the total paid by users in those markets, of which the HIP-3 deployer keeps half. The cumulative notional volume of those markets amounts to $420.686 billion, with $100.017 billion in the last thirty days. Perimeter warning: we do not calculate what share this volume represents of the chain total, as it cannot be verified whether the denominator published by the aggregator includes or excludes HIP-3 markets themselves. The concentration of HIP-3 in a single operator and the peak open interest of $3.2 billion reached in June were documented when examining to what extent Hyperliquid is truly permissionless, and the regulatory fit of its private equity perpetuals in the SEC-CFTC front analysis.

Why is front-end revenue collapsing while HIP-3 grows?

The monthly revenue series published by DefiLlama for each protocol allows for a comparison of the two layers on the same time scale. On the left, the builder code fees for BasedApp—only the main Based application, the narrow perimeter of the previous warning—the second-largest historical builder. On the right, the total fees generated in trade.xyz's HIP-3 markets. Both August 2026 months are partial, with data up to the 3rd.

BasedApp peaked in September 2025 with $6,066,350 monthly and closed July 2026 at $29,686: a 99.5% drop from the high. The slope has been continuous for ten consecutive months, with no month interrupting it except March 2026. Nor does it only affect Based: the entire builder code layer earned 27% less in the last thirty days than in the previous thirty, with drops of 85% for Dreamcash, 82% for Infinex, and 75% for pvp.trade.

The most sober reading of that curve points to points programs. The September 2025 peak coincides with the phase where trading from a third-party front-end accumulated rewards, and a large part of that volume responded to the incentive rather than an interest in trading. When the incentive runs out, the residual volume measures the application's real audience, and in most cases, that audience turns out to be small. Blockworks documented in December 2025 that nearly 40% of Hyperliquid's daily active users traded through third-party interfaces, with a peak above 50% in October of that year; that remains the last known public figure for user distribution, and subsequent revenue suggests the percentage has receded significantly since then.

The trade.xyz curve moves in the opposite direction, and its mechanics explain why. A HIP-3 deployer keeps half of the structural fee of a market that only exists because they created it, without depending on the user approving a revocable surcharge. Their income is tied to the market's own demand—perpetuals on pre-IPO stocks, in their case—rather than retaining traffic against the native interface. In exchange, they put 500,000 HYPE at risk of slashing and assume responsibility for the oracle, a threshold that very few teams have crossed as of August 3, 2026.

Was the Based token already released or is there a pending airdrop?

The idea that Based has a Lighter-style distribution event ahead still circulates. The verifiable fact is earlier: the Based Foundation announced on March 17, 2026, that the initial issuance of the BASED token would take place on March 30, and it did. The total and maximum supply is fixed at 1 billion tokens, and 23.5%—235,400,000 BASED—was unlocked at the event, destined entirely for the community pool. The declared distribution of the remainder places 23.64% in ecosystem and community rewards—a different line item from the community pool despite the similarity in name and figure: the former was fully unlocked at issuance, while the latter is subject to vesting—, 20.36% for investors, 20% for core contributors, 7.5% for the Ethena community (the synthetic dollar protocol), and 5% for seasonal rewards, with three-year vesting.

That 23.5% figure resolves a range that reached the assignment with two incompatible values: BingX listing sheets indicated 24% unlocked at the start, while KuCoin's indicated 36%. The 24% is a rounding of the community pool percentage; the 36% does not correspond to any line in the published allocation table. Based's public documentation does not currently display an accessible tokenomics section—usual GitBook routes return 404s—so the usable reference is the aggregated allocation table—a third-party reconstruction, not an issuer document, including the line item labels—which is 100% internally consistent and coherent with the confirmed issuance date.

Alongside that date, it is worth noting the operational metric. March 2026, the month of issuance, was the only revenue rebound for Based in the entire downward series: $623,425 compared to $444,757 in February. From April onwards, the slope resumes without interruption down to $29,686 in July.

Why does the fee charged by each interface vary twenty-two-fold?

The implicit fee column in the first table contains the most revealing product decision in the sector. pvp.trade captures 0.206% of the volume it routes, while Axiom Pro captures 0.009%: a twenty-two-fold difference within the same protocol ceiling. The 0.10% cap applies per side of the trade, so a builder charging the maximum on both sides can capture nearly 0.2% of the notional counted once; pvp.trade operates right at that limit. At the other extreme, Axiom Pro and Insilico Terminal—0.009% and 0.011%—route massive volumes of professional profiles, where any perceptible surcharge drives the user toward the native interface.

The result is that routed volume is a poor predictor of billing. MetaMask has moved $9.866 billion and has collected $8,340,371; Insilico Terminal has moved $34.167 billion—three and a half times more—and has collected $3,712,578, less than half. The wallet charges 0.085% because its user pays for convenience; the terminal charges 0.011% because its user compares costs with a spreadsheet. Both execute against the same book and with the same depth, yet their revenue per dollar traded is separated by a factor of eight. Anyone evaluating where to trade perpetuals today would do well to look at that surcharge rather than the brand on the screen: the comparison between major perpetuals engines is played out in tenths of a basis point that the front-end can double without the user noticing.

Who is growing now in the distribution layer?

Beneath the surface of the general collapse, there is movement in the opposite direction, coming from generalist wallets arriving late but with an installed base. Trust Wallet earned $181,620 in the last thirty days compared to $11,677 in the previous thirty, and has accumulated $199,017 since integrating: practically its entire history was generated in the last month. Sushi, the DEX aggregator that also operates its own trading interface, went from $55,164 to $156,633 in the same comparison. Invo Perps rose from $263,041 to $441,209 and is now the third-largest builder by 30-day revenue, ahead of names with ten times larger cumulative totals.

The contrast with the case studies from the beginning of the year is instructive. Okto, the wallet that was presented as an example of the model despite running zero-fee campaigns, has accumulated $884,780 but earned $242 in the last thirty days. Dexari, the mobile front-end raised with $2.3 million in funding by former Binance.US executives, has accumulated $617,818 on $2.913 billion in volume and has fallen to $1,744 in the last thirty days. Both were among the cited examples of the model in early 2026, and neither appears among the top fifteen by recent billing anymore.

The operational conclusion for anyone looking at the integration list is that the order by historical cumulative and the order by recent billing no longer coincide. The top ten by cumulative retain 80% of the history but only 52% of the last month's revenue. Turnover is fast because the entry cost is 100 USDC and the exit cost is zero.

What is left for someone wanting to build a front-end on Hyperliquid today?

The August 2026 numbers paint a very different business than the one sold a year ago. The technical barrier remains almost non-existent: two fields in the order payload, 100 USDC in the account, and a user signature. The economic barrier, however, has become the usual one in any app market: 51 of the 102 builders have billed less than $100,000 in their entire history. The mechanism democratizes access to revenue, not the audience that generates it.

Three criteria can be deduced from the series. First, that volume routed during a points campaign measures nothing lasting: BasedApp's 99.5% drop summarizes this. Second, that the ability to set price depends on the type of user and not the product: the twenty-two-fold range in the first table. And third, that those with capital and oracle risk tolerance will find more margin in HIP-3, where the split is 50% on a proprietary market instead of a revocable surcharge on someone else's market, in exchange for 500,000 HYPE at stake.

Hyperliquid's distribution layer has gone in one year from the promise of an open bar to a market with identifiable winners and a long tail that does not monetize. It remains the cleanest example of how a protocol can outsource its interface without giving up the engine, which is why it pays to keep looking at where the fee is paid before deciding which screen to trade from. The revenue fundamentals of the protocol itself follow a different curve, and that distinction explains much of what has been told here.

Sources and links: Hyperliquid Docs — Builder Codes (mechanics, caps, and fill files) · DefiLlama — fees by protocol on Hyperliquid L1 (live query Aug 3, 2026) · DefiLlama — BasedApp (volume and monthly series) · DefiLlama — tradeXYZ (HIP-3 market fees) · DefiLlama open adapter for Hyperliquid (methodology) · CoinGecko Research — Top Hyperliquid Builders (snapshot May 25, 2026) · Blockworks — Hyperliquid: the frontend wars (Dec 4, 2025) · ChainCatcher — Based Foundation confirms March 30 issuance · BASED allocation table (issuance date and 23.5% unlock) · CryptoBriefing — Phantom surpasses $20 million in builder code revenue