Hyperliquid Builder Codes: Pump.fun Wires In as Hyperliquid Holds 56% of Onchain Perp OI

Hyperliquid builder codes are turning wallets and consumer apps into order routers, and the numbers published this weekend show how far that has gone. A Castle Labs report relayed by Wu Blockchain and KuCoin News on October 10 puts Hyperliquid's cumulative protocol revenue above $1.4 billion, more than $1.26 billion spent on open market HYPE buybacks, and Hyperliquid's share of onchain perpetuals open interest above 56%. The same report says Phantom has earned over $25 million through builder codes and MetaMask over $10.5 million.

On the same day, ChainCatcher, citing Wu's Blockchain, reported that a Pump.fun frontend API endpoint was returning Hyperliquid perpetuals data, including HIP-3 stock and commodity markets. For an orderflow trader, the takeaway is simple: more of the flow hitting the Hyperliquid book now arrives through apps that are not Hyperliquid, and that changes who is on the other side of your fills.

What the Pump.fun Endpoint Actually Shows

According to ChainCatcher, the endpoint frontend-api-v3.pump.fun/coins/perps returned data on Hyperliquid perpetual contracts covering BTC, ETH and SOL, plus stock and commodity contracts such as Nvidia, Tesla, gold and crude oil. The fields included contract price, funding rate, open interest, 24 hour volume and the maximum margin multiple for each market. Some of the listed assets come from HIP-3 markets.

The report is careful on one point, and anyone trading on it should be too: it is currently unclear whether Pump.fun plans to launch perpetuals trading, and Pump.fun has not announced a product. A data endpoint is not a trading frontend. What it does show is that a consumer app with a memecoin audience is already pulling the Hyperliquid market list, including HIP-3 equities and commodities, into its own backend.

Lookonchain's Hyperliquid feed logged the same item on October 10 as Pump.fun "reportedly developing" Hyperliquid powered perpetual trading, next to the Castle Labs figures and a scheduled network upgrade, per its Hyperliquid token page. Treat it as a signal under construction, not a launch.

How Builder Codes Route Orderflow to One Book

The mechanism that makes this possible is documented in the Hyperliquid builder codes docs. A builder attaches a builder parameter to orders it sends on a user's behalf, with its address and a fee in tenths of a basis point. Fills sent with a builder code are attributed to that builder, and the fee is processed onchain as part of the normal fee logic.

The constraints are tight. Builder fees are capped at 0.1% on perps and 1% on spot, and on perps the fee applies to both sides of a trade. The user must approve a maximum builder fee from their main wallet, not an API wallet, can revoke it at any time, and can hold at most 10 active approvals. A builder needs at least 100 USDC in perps account value to qualify.

The important part for market structure is what builder codes do not do. They do not create a separate venue or a separate order book. Phantom flow, MetaMask flow and any future Pump.fun flow all rest and match on the same Hyperliquid book as an API market maker. The app owns the user interface and the fee. Hyperliquid owns the matching, margin and liquidation engine.

Phantom, MetaMask and the 56% Onchain Open Interest Share

The Castle Labs numbers give a sense of scale. Phantom's $25 million and MetaMask's $10.5 million in cumulative builder revenue, per TokenPost, only come from fees on real fills. At the 0.1% maximum perps fee, every dollar of builder revenue implies at least a thousand dollars of routed notional, so Phantom's $25 million implies at least $25 billion of routed notional, and more if it charges below the cap.

The daily picture has been moving too. FinanceFeeds, citing Blockworks Research data shared by Wu Blockchain, reported a record $984 million in daily perps volume routed through builder codes on August 21, 2026, with about $782,000 in combined builder revenue that day. MetaMask took about $143,000 and Phantom about $114,000.

For context on direction, Blockworks Research wrote in December 2025 that nearly 40% of Hyperliquid daily active users traded through third party frontends, while over 90% of volume still came from the native frontend. Users moved first, volume followed slower. The 56% onchain open interest share, cited by Coin Turk from the same Castle Labs work, says that more than half of onchain perp open interest sits on Hyperliquid, whatever app the orders came through.

There is a revenue loop behind this. Coin Turk reports $72.2 million of HYPE bought on the open market over the prior 30 days, citing the buyback data, and says repurchased tokens sit with the Assistance Fund rather than being burned. More routed flow means more fees, and more fees mean a larger mechanical bid for HYPE.

What Wallet and App Flow Changes in the Orderbook

The open question is what kind of taker that flow is. Builder fills are attributed per builder in daily files that Hyperliquid publishes, per the docs, so the composition can be measured rather than guessed. Until it is, treat a new frontend as a change in the mix of aggressors, not as extra liquidity. If a memecoin audience ever gets one tap access to HIP-3 Nvidia or crude oil perps, those books will see takers they do not see today.

That matters for three readings orderflow traders rely on:

• Cumulative volume delta. CVD measures net aggressive buying minus aggressive selling. A run in CVD built from many small market orders reads differently from one built by a single large wallet, and the follow through usually differs too.
• Funding. A one sided push into a thin HIP-3 market pays funding to whoever takes the other side. Watching funding against open interest shows whether the crowd is building positions or rotating.
• Liquidation density. When many accounts enter near the same price with similar margin, liquidation levels stack in the same zone. That stacking is where cascades start, whichever app the orders came from.

None of this requires Pump.fun to launch anything. Phantom and MetaMask already route flow this way today, and the Castle Labs numbers show it is not marginal.

Tracking Routed Flow on Buildix

The practical job is separating where flow is coming from and what it does to the book. On Buildix, the screener covers Hyperliquid perps including HIP-3 markets, so a sudden open interest and funding move on a stock or commodity perp stands out next to the crypto majors. The whale tracker handles the other end, separating a few large wallets from a crowd of small ones when CVD starts to run.

Hyperliquid's moat is not a frontend. It is one book that every frontend has to route into. As more apps plug in, the order book gets more heterogeneous, and reading who is aggressing becomes more valuable than reading how much.