GMTRENCHOR

GET /x402/hl/funding

Funding, open interest and premium for every perp on Hyperliquid, across every venue at once.

1¢ per call — 10000 units of USDC, payable on Arc, Base, Arbitrum, HyperEVM. Paid to 0xf01BccA948bB974D6B8CEb53705524bd9E529D52 through https://gateway-api.circle.com/v1/x402.

Hyperliquid's own API answers one venue at a time and keeps no history of open interest at all. This is the whole board in one call: every asset on the native exchange and on each builder-deployed venue, with its funding rate, open interest in dollars, premium, mark and oracle price, and the day's volume — sampled by us every few minutes and stated with the time it was read.

Answer

assets: one row per asset, with a row per venue inside it; venues: the venues sampled and how many markets each carries

Status codes

GET /x402/hl/carry

The widest funding spreads between venues right now, above an open-interest floor.

5¢ per call — 50000 units of USDC, payable on Arc, Base, Arbitrum, HyperEVM. Paid to 0xf01BccA948bB974D6B8CEb53705524bd9E529D52 through https://gateway-api.circle.com/v1/x402.

The same asset funds at different rates on different Hyperliquid venues, and the gap is a carry an agent can take by being long on one and short on the other. No single Hyperliquid call returns it, because no single call sees two venues. Sorted widest first, and floored by open interest so the list is tradeable rather than dust.

Answer

carry: asset, the cheap venue and the dear one, both rates, the spread, and the open interest behind each leg

Status codes

GET /x402/hl/asset

One asset across every venue that lists it, with book imbalance.

1¢ per call — 10000 units of USDC, payable on Arc, Base, Arbitrum, HyperEVM. Paid to 0xf01BccA948bB974D6B8CEb53705524bd9E529D52 through https://gateway-api.circle.com/v1/x402.

Every venue's funding, open interest and price for one coin side by side, plus the resting-size imbalance at the top of its book where we sample it. Open interest is unsigned and says nothing about which side is crowded; the book imbalance is the part that does.

Query

Answer

venues: one row per venue; book: bid share of resting size at the top of the book, and the depth it was measured over

Status codes

GET /x402/hl/history

Hourly funding, open interest and book imbalance since 17 August 2026.

$19.00 per call — 19000000 units of USDC, payable on Arc, Base, Arbitrum, HyperEVM. Paid to 0xf01BccA948bB974D6B8CEb53705524bd9E529D52 through https://gateway-api.circle.com/v1/x402.

Hyperliquid publishes no history for open interest and none for the order book, so a sample not taken is gone. This is ours, taken every few minutes since 17 August 2026 and folded to the hour: funding, open interest and book imbalance per asset per venue. It states how many of the possible hours are actually present, because a gap is a fact about the data and not something to paper over. This file carries the fifty busiest assets; /x402/hl/history/full carries every asset on every venue.

Answer

hours: one row per hour per asset; coverage: hours present against hours possible, and the first and last

Status codes

GET /x402/hl/history/full

Every asset on every Hyperliquid venue, hourly since 17 August 2026: funding, open interest and book imbalance.

$129.99 per call — 129990000 units of USDC, payable on Arc, Base, Arbitrum, HyperEVM. Paid to 0xf01BccA948bB974D6B8CEb53705524bd9E529D52 through https://gateway-api.circle.com/v1/x402.

The complete series rather than the busiest fifty: every asset on the native exchange and on all ten builder-deployed venues, hour by hour since 17 August 2026, with funding, open interest in dollars, the mark, and book imbalance where the book was sampled. Hyperliquid itself serves no history of open interest and none of the order book, so every such series on the market is somebody's own recording, ours included. Others do sell recordings of the builder venues — checked 18 September 2026 — and they charge for the right to use them: commercial rights start at $3,588 a year with CoinGlass, which carries no builder venues at all, and $4,992 a year with the nearest vendor that does. This is $129.99 once, commercial use included, and the derived book-imbalance series is not sold by anyone we could find. Same shape as the standard history, so one parser reads both. The response is gzip-encoded JSON and an ordinary HTTP client inflates it transparently. It states how many of the possible hours are present.

Answer

hours: one row per hour per asset per venue, every asset; coverage: hours present against hours possible, and the first and last

Status codes

GET /hl/builders

What each Hyperliquid builder actually charges, measured from their own published fills.

Free. No payment, no wallet.

Hyperliquid publishes every builder's fills at a URL keyed only by address and date. This is what those files say each builder took: fills, volume, fee collected and the effective rate in basis points, by day. It is what a competitor charges in practice rather than what a dashboard claims. Free.

Answer

builders: address, label, and per-day fills, volume, fee and basis points

Status codes

GET /x402/arc/rates

What USDC and EURC earn and cost to borrow on Arc, every market, right now.

1¢ per call — 10000 units of USDC, payable on Arc, Base, Arbitrum, HyperEVM. Paid to 0xf01BccA948bB974D6B8CEb53705524bd9E529D52 through https://gateway-api.circle.com/v1/x402.

Read straight off the lending contracts on Arc rather than from an index: Aave and Morpho hold almost all of the chain between them. Each market gives supply and borrow rates, utilisation, available liquidity and the collateral behind it. Note that Aave on Arc is version 4, a Hub-and-Spoke design with no Pool and no aTokens, so an address book written for version 3 will not find it.

Answer

assets: per asset, the best supply rate and the size of the pool offering it; markets: every market with its rates, utilisation and liquidity

Status codes

GET /x402/arc/best-yield

Where a given size of USDC earns most on Arc once the deposit's own effect is counted.

5¢ per call — 50000 units of USDC, payable on Arc, Base, Arbitrum, HyperEVM. Paid to 0xf01BccA948bB974D6B8CEb53705524bd9E529D52 through https://gateway-api.circle.com/v1/x402.

A quoted rate is what a pool pays as it stands, and depositing into a small one dilutes it. This answers what a specific size would actually earn: every market ranked by the rate after the deposit lands, with the pool's current depth beside it. Measured on Arc on 18 September 2026, the best quoted USDC rate was 6.26 percent in a market holding eleven thousand dollars — the same size that earns 0.007 percent once ten million is added to it.

Query

Answer

best: the market that wins for that size; options: every market, with the rate before and after the deposit

Status codes

GET /record

The scanner's own scoreboard: what it called, what happened, and what its research engine has refuted.

Free. No payment, no wallet.

Every call opened at a price, tracked and closed at a fixed horizon by the same code, with nothing edited and nothing removed. Rates come with their denominators and with the count of calls the record could not vouch for. It carries the lift against pools that traded and were never called, which is the only question that matters about a scanner; the outcome of every score band, including the bands that do not separate; and the theses the research engine has refuted, which is most of them so far. Free, because what is sold here is the coverage and the data rather than the tip, and a record that only showed the parts where we look good would not be a record.

Answer

chains: per chain: calls, closed, unverified, outcomes, rates, medians, and the lift against pools never called; scoreBands: outcome by score, so the score answers for itself; labs: the research engine per chain: what it has confirmed, refuted and still holds open, with out-of-sample night counts, kept apart because a thesis that holds on one chain and fails on the other is a finding; caveats: what these numbers cannot tell you

Status codes

GET /x402/calls

Every call open right now on Arc, with what was known when it was opened.

3¢ per call — 30000 units of USDC, payable on Arc, Base, Arbitrum, HyperEVM. Paid to 0xf01BccA948bB974D6B8CEb53705524bd9E529D52 through https://gateway-api.circle.com/v1/x402.

A record of what a scanner flagged and what has happened since. Each row carries the entry price, the score and the reasons the gate fired, the pool's liquidity and fully diluted value at that moment, how old the pool was, the return so far, the sell-test verdict honeycheck had already published, and the Hyperliquid market at the moment the call was opened. A call is opened at the pool's price, not at a fill, and nothing here is advice or a recommendation. Arc only: our own scoreboard says the same gate does not work on Robinhood, and the free record at /record shows the figures.

Answer

open: one row per open call: entry, score, why, liquidity and value at call, age, last price, return so far, sell test, market at open; why: why this feed is Arc only, in one sentence, with the lift figures behind it

Status codes

GET /x402/calls/closed

Every call that has ended, with its outcome where the prices are ours to sell. The labelled set for a backtest.

$49.00 per call — 49000000 units of USDC, payable on Arc, Base, Arbitrum, HyperEVM. Paid to 0xf01BccA948bB974D6B8CEb53705524bd9E529D52 through https://gateway-api.circle.com/v1/x402.

What a backtest of a launch gate actually needs and cannot rebuild afterwards: a claim made at a time, and what happened next. Every call ever opened and since closed, on Arc and on Robinhood, graded by the same code at a fixed horizon and never edited. Read `provenance` before parsing. Arc rows are complete — we read that chain ourselves, block by block — and carry the entry, the score and reasons, liquidity and value at the call, the outcome, the close and best returns, minutes to the best print, whether the pool was pulled, and the Hyperliquid market at that moment. Robinhood rows carry the claim, the time, the score, the reasons and the market, but not the prices or outcomes: every observation of that chain came from Dexscreener and GeckoTerminal, whose terms forbid redistributing or deriving from their data, so those columns are null and each row names them. The rows stay rather than being dropped, because a record showing only the chain where the scanner looks good would not be a record. The response is gzip-encoded JSON and an ordinary HTTP client inflates it transparently.

Answer

closed: one row per closed call; complete where we observed the chain ourselves, claim-only where we did not; provenance: who observed each chain, how many rows are complete, and which columns are withheld and why; counts: closed by chain, how many the record can vouch for, and how many carry prices

Status codes

GET /x402/context

The Hyperliquid market now and a day ago: funding, open interest, book imbalance, for four assets.

1¢ per call — 10000 units of USDC, payable on Arc, Base, Arbitrum, HyperEVM. Paid to 0xf01BccA948bB974D6B8CEb53705524bd9E529D52 through https://gateway-api.circle.com/v1/x402.

The backdrop, for a cent, without buying the whole funding board. Bitcoin, Ether, HYPE and Solana on the native exchange at the newest hour we folded, beside the same hour a day earlier, with the change in each. It also gives the widest funding gap between any two venues carrying each asset, which is the cheapest read on whether a carry exists at all. When the hour a day ago was never sampled, every change column is null and the answer says so rather than quietly comparing against a nearer hour. This is the same block each call in /x402/calls carries.

Answer

assets: per asset: funding, open interest, bid share and mark, now and a day ago, with the change; dayAgoPresent: whether the comparison hour was actually sampled

Status codes

GET /x402/hl/carry/stats

How long funding spreads between Hyperliquid venues actually last, from our whole record.

5¢ per call — 50000 units of USDC, payable on Arc, Base, Arbitrum, HyperEVM. Paid to 0xf01BccA948bB974D6B8CEb53705524bd9E529D52 through https://gateway-api.circle.com/v1/x402.

A spread you can see is a snapshot, and a snapshot cannot be sized: a gap open for one hour in fifty and a gap open for three weeks look identical at the moment you read them. Per asset and venue pair, over every hour we have sampled since 17 August 2026: how many hours we hold both legs for, the share of them the spread stood above each level, how long an episode above a level ran in consecutive hours, and how often the spread changed sign inside a day, which is what turns a carry into two fees and no position. Both legs are floored by open interest so the list is sizeable rather than dust, and the floor is published. Nothing is annualised: funding is an hourly rate and multiplying it by a year states a rate that has never held for a week. Hyperliquid keeps no history of funding by venue, so this cannot be recomputed from any public source.

Answer

pairs: per asset and venue pair: hours observed, share above each level, episode lengths, sign reversals by day, and the spread now; window: the first and last hour folded, and how many hours that is

Status codes

GET /x402/arc/token

One Arc token's first hour, frozen: who bought, how concentrated, and what happened after.

5¢ per call — 50000 units of USDC, payable on Arc, Base, Arbitrum, HyperEVM. Paid to 0xf01BccA948bB974D6B8CEb53705524bd9E529D52 through https://gateway-api.circle.com/v1/x402.

The Arc store keeps the swap tape for 48 hours and then prunes it, so a pool's first hour is knowable for two days and then gone from the public node for ever. This is ours, written down while it existed: trades, buys and sells, distinct buyers and sellers, dollars through the pool, the largest single buy, what share of the hour's money was three wallets, and the price at ten, thirty and sixty minutes. Beside it goes what was knowable at the time and only that — the sell-test verdict already published, the holder count and concentration, whether our scanner called it — then the outcome once its horizon passed, the Hyperliquid market at the moment it started trading, and the deployer with a summary of what else they have launched. Whether the token can be sold right now is honeycheck's measurement, made with its own money; the answer points at it and does not repeat it.

Query

Answer

firstHour: the pruned window, frozen: trades, buyers, concentration, dollars, prices at ten, thirty and sixty minutes; knownEarly: sell test, holders and whether it was called, each with the time it was known; outcome: returns once the horizon passed, null before it; deployerHistory: what else the same address has launched, where we have attributed it

Status codes

GET /x402/arc/deployer

What one Arc deployer has put out, and what became of it.

5¢ per call — 50000 units of USDC, payable on Arc, Base, Arbitrum, HyperEVM. Paid to 0xf01BccA948bB974D6B8CEb53705524bd9E529D52 through https://gateway-api.circle.com/v1/x402.

Pools attributed to an address by reading each pool's own creation transaction off Arc. How many we hold, when they were first and last seen, the share that anyone traded, the share that died and the share that doubled once their horizons passed, the median dollars through the first hour and the median share of it taken by three wallets, and the sell-test verdicts counted one by one rather than collapsed. Every rate states the denominator it was computed over, because the attribution pass is paced and budgeted and is therefore partial by design: this is what we could attribute, never a claim about everything an address has launched. A verdict of TAXED means the money came back short, not that it did not come back, and it is counted separately for that reason.

Query

Answer

traded: how many of the attributed pools anyone traded, and the share; graded: died, doubled and median returns, over the pools whose horizon has passed; sellTested: verdicts by name, with the two that mean the money did not come back counted apart; recent: the fifty most recent, each with its first-hour size and outcome

Status codes

GET /x402/panel

One frozen day of every Arc launch, first hour and all, as a single file.

$9.00 per call — 9000000 units of USDC, payable on Arc, Base, Arbitrum, HyperEVM. Paid to 0xf01BccA948bB974D6B8CEb53705524bd9E529D52 through https://gateway-api.circle.com/v1/x402.

A whole day of the launch panel: one row per pool that started trading against USDC on Arc that day, with its first hour frozen from our own swap tape, what was knowable at the time, the outcome once every horizon inside the day had closed, the Hyperliquid market at the moment it started, and the deployer where we attributed it. A day is offered only once every 24-hour outcome inside it has closed, which is two days after it began; /panel/days lists which are available and which are still filling. A row still pending after that never printed a price again, and the count is published rather than dropped. The response is gzip-encoded JSON and an ordinary HTTP client inflates it transparently.

Query

Answer

rows: one row per pool that started trading that day; counts: how many rows carry a first hour, an outcome, a sell test, a holder reading, a call and a deployer

Status codes

GET /panel/days

Which days of the Arc launch panel can be bought, and which are still filling.

Free. No payment, no wallet.

The index for /x402/panel. Each available day comes with how many pools it holds, how many outcomes are complete and how many are still pending, and how large the file is, so a buyer knows what they are buying before they pay for it. Days that exist but are not yet offered are listed too, with the reason. Free.

Answer

available: the days that can be bought, with their counts and size; notYet: days that exist but are still filling, and why

Status codes

GET /x402-meter.json

Circle Gateway activity on Arc, measured at the contract every x402 payment settles through.

Free. No payment, no wallet.

Gateway is where x402 payments on Arc settle, ours included, and its logs are public. This counts spends out of Gateway balances, deposits into them, and batch settlements, kept apart because they are different things. Nothing here is called a payment: a spend covers cross-chain USDC transfers too, and the destination breakdown shows how much left the chain. Read the concentration figure before reading the total as a market — one address has been most of it. Free.

Answer

counts: spends, deposits, batch settlements, and the distinct addresses behind them; byDestination: spends and USDC by destination domain, so cross-chain transfers are visible; concentration: the share of the total one address accounts for; days: one row a day; topSpenders: addresses by USDC spent, with when each was first seen

Status codes