Free API for perp market data

Perp DEX volume overview (24h/7d/30d, change %, per protocol).

Fields: protocols[], dataStatus, generatedAt, upstreamTimestamp, schemaVersion, sources, ETag

Live perp execution-cost ladder from order books and supported oracle/pool models. Total cost is fee + half-spread + size-dependent impact beyond the best quote. A venue that cannot fill the order publishes totalBps null and keeps its partial-fill numbers under partial. depthConfidence low marks displayed depth the plausibility gate could not confirm; such rows rank after confirmed rows. An oracle venue has no book: GMX publishes the size-dependent GMX v2 position impact read from the on-chain DataStore through the GMX Reader, in priceImpactBps, with slippageModel onchain-position-impact-v2 and slippageConfidence medium; slippageSourceUrl names the contract that was read. Caching: a healthy answer carries Cache-Control public, max-age=10, s-maxage=25, must-revalidate, so one measurement serves every reader for the engine TTL; partial and stale answers use max-age=0, s-maxage=5, must-revalidate so readers see a completed background refresh before its fresh window expires. X-Cache states where the answer came from: HIT (this instance), SHARED (the shared result store, inside the TTL), STALE-SHARED or STALE (a labelled older result, with staleAgeSeconds, refreshed behind the response), MISS (measured now) or UNAVAILABLE (nothing inside the stale cap).

Fields: asset, sizeUsd, side, results[].{exchange,feeBps,feeBpsMin,feeBpsMax,feeModel,slippageBps,halfSpreadBps,spreadBps,totalBps,midPrice,vwap,sufficient,fillStatus,depthFetchedUsd,depthConfidence,depthConfidenceReason,rankingEligible,partial,quoteStatus,maxExecutableSizeUsd}, coverage, referenceMidSource, meta, dataStatus, generatedAt, schemaVersion, sources, ETag

Live spot execution-cost ladder using venue order books, pair-specific fees, spread, VWAP impact and explicit depth coverage.

Fields: asset, sizeUsd, side, results[].{exchange,baseAsset,quoteAsset,feeBps,slippageBps,halfSpreadBps,spreadBps,totalBps,midPrice,vwap,sufficient,fillStatus,depthFetchedUsd,depthConfidence,depthConfidenceReason,partial}, coverage, meta, dataStatus, generatedAt, schemaVersion, sources, ETag

Direct DEX market enrichment: OI, volume, funding, prices, protocol stats and venue coverage. fetchedAt is the time the upstreams were read — cross-venue funding predictions are judged against it, not the serve clock.

Fields: protocols[], crossVenueFunding, fetchedAt, coverage.{requested,succeeded,failed,rejected,retired}, dataStatus, generatedAt, upstreamTimestamp, schemaVersion, sources, ETag

Aggregate daily perp-volume time series (DefiLlama) plus total open interest per day from our own snapshots. oi is absent on days with no capture — never zero-filled.

Fields: chart[{date,volume,oi?}], oiCoverage.{daysWithOi,days,note}, dataStatus, generatedAt, schemaVersion, sources

24h volume per venue from direct venue-reported perp tickers (CEX) and perp-DEX protocol APIs, with top symbols. Each row carries venueType so consumers can split CEX from DEX. exchanges[] is sorted by volume24h and holds the CEX rows plus Hyperliquid and dYdX (both venueType dex); every other perp DEX sits in dexVenues[], and dexTotalVolume is the sum of dexVenues[] only. Two hold-out rules keep a row visible while keeping its figure out of totalVolume. spikeFlagged[] names venues whose reported 24h volume is above 3x their own 7-day median (row carries volumeSpike true; the hold-out ends when the print holds). unverified[] names venues whose reported turnover PerpFinder cannot verify against an open-interest feed or an order book of matching size (row carries volumeUnverified true; the hold-out is permanent until an editorial review removes the entry). heldOut[] is the union, one {name, reason} per excluded venue, and each row carries heldOut set to the string unverified, the string spike, or null. change_1d is the day-over-day change of our captured snapshot total between the two newest calendar-adjacent captures, with every heldOut venue removed from both days, so it sits on the same basis as totalVolume; null when the two captures are not one day apart.

Fields: exchanges[].{name,volume24h,symbolCount,topSymbols[],venueType,volumeSpike?,volumeUnverified?,heldOut}, totalVolume, spikeFlagged[], unverified[], heldOut[].{name,reason}, dexVenues[], dexTotalVolume, noFeed[], change_1d, failed[], venueSlugs, updatedAt, dataStatus, generatedAt, upstreamTimestamp, schemaVersion, sources, ETag

Open interest per venue and the aggregate total. byExchange[] keeps the historical CEX list (plus Hyperliquid and dYdX); other perp DEXes are additive under dexVenues[]. Three totals, all in USD: totalOI is the sum of byExchange[] only (unchanged for existing consumers), cexTotalOI is the sum of the byExchange[] rows tagged venueType cex, and totalOIAllVenues is totalOI + dexTotalOI. coverage.cex and coverage.dex count the venues in THIS response; the /open-interest page replaces its initial server snapshot with this feed on each successful refresh.

Fields: totalOI, cexTotalOI, totalOIAllVenues, byExchange[].{name,oi,venueType}, dexVenues[], dexTotalOI, failed[], venueSlugs, coverage.{cex,dex,failed,noFeed,rejected,symbolCoverage}, updatedAt, dataStatus, generatedAt, upstreamTimestamp, schemaVersion, sources, ETag

DEX long vs short open-interest split per protocol.

Fields: protocols[], totalLong, totalShort, updatedAt, dataStatus, generatedAt, schemaVersion, sources

Live funding-rate matrix: asset × exchange (normalized 1h). Missing OI/price are null with per-venue field support and explicit feed coverage. Venues that report a per-market interval (Binance, Bybit, OKX, Bitget, MEXC, Gate.io, BingX, KuCoin, Phemex, HTX, BloFin, Orderly, Aster) keep the raw rate and that interval; a market with no reported interval gets no rate. When only the interval request fails, the last reported interval of the market (at most 24 h old) is used and marked intervalSource "last-known" with intervalAgeMs. Optional asset/venue/limit filters narrow the matrix; a venue filter recomputes the per-row aggregates.

Fields: rows[].exchanges.{rate1h,rawRate?,intervalHours?,intervalSource?,intervalAgeMs?,oi,price}, exchanges[], coverage, updatedAt, meta.fieldSupport, meta.venueSlugs, dataStatus, generatedAt, upstreamTimestamp, schemaVersion, sources, ETag

Protocol fees (DEX, 24h/7d/30d per protocol) — aggregator-reported protocol fees, one row per Derivatives protocol. Fees paid and protocol-retained revenue are different measures. It carries no CEX rows: a CEX figure on /fees/generated is that venue’s 24h volume from /api/data/volume multiplied by a maker/taker rate. The volume rule therefore decides the CEX estimate too — a venue that /api/data/volume holds out, either as volumeSpike (24h volume above 3x its own 7-day median) or as volumeUnverified (reported turnover we cannot verify), keeps its row and its chip on /fees/generated, and stays out of the 24h fee total and the share bars, exactly as it stays out of totalVolume.

Fields: protocols[], dataStatus, generatedAt, upstreamTimestamp, schemaVersion, sources, ETag

Liquidations 24h/4h/1h, long vs short split. Binance and Bybit come from our own capture of their forced-close WebSockets; OKX, Gate.io, HTX and GMX come from their public REST feeds. meta.coverage states how much of the last 24 hours the capture actually covered and which venues have no continuous capture. Captured-venue windows are summed from stored five-minute slots, so a window edge includes the slot it falls inside.

Fields: events[], summary, meta.{venues,missingVenues,windowHours,coverage.{status,collector24h,collectorPerVenue,venuesWithoutCollector,collectorSince}}, updatedAt, dataStatus, generatedAt, upstreamTimestamp, schemaVersion, sources

Hourly liquidation history for Binance and Bybit from our own socket capture. Ledger v2 uses Bybit position sides and observed heartbeat windows; v1 remains separate. An hour with no capture returns null totals, never 0, and every bucket carries its coverage share.

Fields: collectorVersion, methodologyNote, series[].{t,longUsd,shortUsd,events,coverage,byVenue}, coverage.{fraction,perVenue,hoursWithCoverage,hoursRequested}, collectorSince, observedTotalUsd, hours, venues[], updatedAt, dataStatus, generatedAt, upstreamTimestamp, schemaVersion, sources

BTC/ETH options market state: mark IV per strike/expiry, ATM term structure, put/call, Deribit DVOL (attributed), expiries.

Fields: status, updatedAt, assets.{BTC,ETH}.{asset,venue,status,updatedAt,spotIndex,dvol,summary,termStructure[],skew[],skewExpiry,expiries[]}, dataStatus, generatedAt, schemaVersion, sources

Self-collected options time series (no synthetic backfill); allowlisted metrics; maturity + methodologyVersion in every response.

Fields: series[], maturity, methodologyVersion, firstObservedAt, observations, start, end, dataStatus, generatedAt, upstreamTimestamp, schemaVersion, sources, ETag

Reported vs observable CEX volume signals with separate data-confidence scoring. sweepsInWindow counts the last sweepWindowDays UTC days, not the lifetime total — use volume-quality-history for a longer count. unassessedTopVenues names the top-20-by-reported-volume venues that no sweep measures, with the reason.

Fields: sweep.venues[], maturity, methodologyVersion, datasetStart, firstSweepAt, sweepsInWindow, sweepWindowDays, sweepCountScope, unassessedTopVenues[].{venue,reportedVolume24hUsd,rankByReportedVolume,reason}, unassessedRankSource, dataStatus, generatedAt, schemaVersion, sources

RWA perp markets: stocks / forex / commodities across venues. platforms[].funding is an hourly rate fraction; null when the venue or its interval is unknown.

Fields: markets[], updatedAt

Committed ESMA MiCA register snapshot joined to PerpFinder venue status, service permissions, EU-derivatives scope and fee context.

Fields: register, scope, coverage, authorized[], tracked[], dataStatus, generatedAt, upstreamTimestamp, schemaVersion, sources, ETag

Binance USDT-M perpetual premium index — (mark − index) / index, single venue. indexPrice is Binance’s own index, not an independent spot reference.

Fields: entries[].{symbol,venue,markPrice,indexPrice,premium,fundingRate}, avgPremium, venue, updatedAt, dataStatus, generatedAt, schemaVersion, sources

Fear & Greed index (alternative.me, attributed). current is null when the index is unavailable.

Fields: current.{value,classification,timestamp}, history[], updatedAt, staleSince, dataStatus, generatedAt, upstreamTimestamp, schemaVersion, sources

Self-collected funding history: aggregates, per-venue sweeps, or the append-only realized-settlement ledger. Rate types never mixed; no synthetic backfill. Sweep history defaults to one joined series: v2 sweeps up to the first v3 sweep (2026-09-23), then v3 up to the first v4 sweep (2026-09-25), then v4 (selectionMode "joined", methodologyVersions [2,3,4], seriesNote, excludedFields). Hourly rates are comparable across the join; v2 Aster and Crypto.com rawRate/rawIntervalMinutes are null; v2/v3 Hyperliquid predicted nextSettlementAt is the venue-reported start of the running hour, v4 points at the next settlement. methodologyVersion=1, 2, 3 or 4 selects one stored version only. Each version has its own dataset start.

Fields: series[]/entries[], coverage, maturity, methodologyVersion, methodologyVersions, selectionMode, seriesNote?, excludedFields?, datasetStart, dataStatus, generatedAt, schemaVersion, sources, ETag

Funding prediction accuracy per venue and asset: stored predicted rates matched to the append-only realized-settlement ledger over one shared window per venue and asset (after the first usable stored prediction, up to the furthest targeted settlement, inside a 30-day read window). Accuracy (MAE and bias in bps per hour of the hourly-normalized rate, sign hit rate, median lead) is returned only when the publication gate passes (>=3 venues x >=30 settlements x >=7 days, <=30% missing); otherwise publication "withheld", accuracy null and the CSV returns 409. Hyperliquid v1-v3 predictions carry a labelled +1 interval settlement-time correction (horizonCorrection).

Fields: publication, gate.{ready,missing}, requirements, readWindow, methodologyVersions, method[], units, readiness[], accuracy[]|null, csv|null, dataStatus, generatedAt, schemaVersion, sources, ETag

Volume-quality sweep history per venue (reported vs observable signals over time). Measurements, never accusations.

Fields: series[], maturity, methodologyVersion, datasetStart, updatedAt, dataStatus, generatedAt, schemaVersion, sources

Perpetual volume ranking of CEX and DEX venues, reported and adjusted. Adjusted is the reported figure with held-out venues removed (unverified tier, 3x spike rule, failed depth test); it is not a modelled estimate. Each held-out venue keeps its reported figure and a dated reason. view=share gives the daily DEX share, raw and adjusted; view=share-monthly the volume-weighted monthly share. date pins one stored day file for citation. Measurements, not accusations.

Fields: ranking.{window,dates,capturedAt,reported,adjusted,gapUsd,gapShare,venues[].{name,type,reportedUsd,adjustedUsd,reportedRank,adjustedRank,holdOuts[].{rule,reason},spikeMultiple,depth}}, share[], monthly[], dayFile, permalink, depthSweepAt, datasetStart, dataStatus, generatedAt, upstreamTimestamp, schemaVersion, sources

Per-venue daily volume/OI history from the committed snapshot series (since 2026-06-05) — venue-reported figures, normalized; null never zero-filled. The window is a calendar window, and every day with no capture is listed.

Fields: series[].{date,observedAt,backfilled,volume24h,volume7d,volume30d,openInterestUsd,perpPairs}, venueSlug, datasetStart, windowStart, windowEnd, windowDays, missingDays, gaps[], missingDates[], note, updatedAt, dataStatus, generatedAt, schemaVersion, sources

Builder fees that wallets and trading apps add to Hyperliquid perps fills, summed from the public Hyperliquid builder-fill files over the newest stored window of up to 7 UTC days. Each row gives the app, its builder addresses, notional, builder fees, the added fee in bps, and the base-tier taker total against a direct Hyperliquid order on $10,000. The same rows as the /trading-apps table. A missing value is null in JSON and empty in the CSV.

Fields: window.{from,to,dates}, directTakerBps, costModel, rows[].{app,builderAddresses,daysCaptured,fills,notionalUsd,builderFeesUsd,addedFeeBps,takerTotalBps,taker10kUsd,premiumOverDirectPct}, updatedAt, dataStatus, generatedAt, upstreamTimestamp, schemaVersion, sources

Execution-cost history by methodology version. While the active version has fewer than 7 days of stored sweeps, the default merges every stored version (selectionMode "merged"); each point still carries its own methodologyVersion and the boundary sweeps are flagged. methodologyVersion=active is the strict active-only series. Use endDate to page an older window.

Fields: series[].{attemptId,ingestedAt,collectedAt,methodologyVersion,versionBoundary,side,feeType,feeTierAssumption,engineDataStatus,coverage,venues[].{halfSpreadBps,priceImpactBps,quoteStatus,slippageModel,executionModel,fillStatus,depthConfidence,depthFetchedUsd}}, lastSampleAgeMinutes, collectorStale, selectionMode, methodologySelectionNote, activeMethodologyMinDays, selectedMethodologyVersions, availableMethodologyVersions, versionBoundaries, globalDatasetStart, datasetStart, maturity, maturityByMethodology, venueCoverage, updatedAt, generatedAt