LiquidVision — Crypto Derivatives Data

ONE CALL instead of five: ask in plain words ('where is leverage stacked on BTC', 'why did ETH drop', 'is SOL funding crowded', 'biggest whales near liquidation') and get answer: short statements with the numbers in them, facts: the same numbers as JSON, sources: which endpoints were read and how old their data is, caveats. The question is read for its topics (liquidations, funding, open interest, whales, levels, sentiment; English or Russian) and only those endpoints are fetched. No model involved: facts, not opinion, so you do the judging. Free, no key. For the base rate of the current state call get_precedents; for a narrative with an LLM call explain_market.

[receipt] A SIGNED snapshot: the wrapped endpoint's JSON plus an attestation (SHA-256 of the canonical JSON, server time, Ed25519 signature, key id). Anyone can verify it offline with the public key at /api/v1/attest/key, so an agent acting for someone else can prove what the exchanges reported when it decided. Free snapshots and conclusions only (no metered history). It certifies what LiquidVision served, not that the exchange was right. Priced: $0.05 per call (x402), 10 prepaid credits, or a key's daily quota; the public key at /attest/key is free.

[payment] How an agent with a wallet buys a prepaid API key by itself, with ONE x402 USDC payment and no human, email or account. Returns the exact payment requirements (price, asset, payTo, network, the resource URL) for GET https://liquidvision.app/api/v1/credits/buy?usd=N. Send that GET with a PAYMENT-SIGNATURE (x402 v2) or X-PAYMENT (v1) header from your wallet and the answer is a key of tier 'prepaid' with the credits ($0.005 each; a row costs 1, a 12-venue position check 4, precedents or an attestation 10, a rule test 20); the same wallet always tops up the same key. Credits are spent per call by price class; every conclusion and live snapshot stays free. This tool itself moves no money.

[measured + modeled] Your own position against the market in one call: the liquidation price on each venue from THAT venue's maintenance-margin tier at this notional (read from the exchanges' public tables, 'estimated' where none is published), the distance from the current mark, how much modeled CEX leverage and how much MEASURED Hyperliquid leverage sits between the mark and your liquidation price on your side, the funding cost per day at the venue's live rate, and how often in our hourly archive price moved at least that far against your side within 24 h and 72 h (a frequency, not a forecast). Assumes isolated margin and ignores fees, so the real level is slightly nearer the entry. ONE venue (venue= given) is free; the 12-venue comparison is an answer we price: $0.02 per call (x402), 4 prepaid credits, or a key's daily quota.

Subscribe an HTTPS URL to push alerts instead of polling (Pro: 3, Bot: 20 subscriptions). type='condition': YOUR OWN trigger on one symbol — fires when metric op threshold becomes true (edge-triggered: once per crossing, again only after it was false). metric = funding_apr_pct (median annualised funding across venues) | oi_chg_1h_pct | oi_chg_24h_pct (Binance+Bybit+OKX, in coins) | liq_1h_usd | liq_24h_usd | price_chg_1h_pct | price_chg_24h_pct (Binance mark) | long_pct (mean long share of accounts) — the SAME definitions test_rule backtests, so test a rule first and alert on it after; op = gt | lt. Example: symbol='BTC', metric='funding_apr_pct', op='gt', threshold=50. The payload carries the value, the threshold and every other metric as context. Checked every 60 s. type='position': YOUR OWN position (symbol, side, entry, leverage, notional_usd, venue) — fires when it is within max_dist_pct of its liquidation price, computed every 60 s from the venue's maintenance-margin tier and the live mark (same numbers as check_position); again only when it gets a whole percent closer. type='cascade': a liquidation cascade just ended, over min_usd and x_normal times the coin's normal minute. type='hl_near_liq': a Hyperliquid whale position over min_usd within max_dist_pct of its liquidation price. type='cluster_sweep': price reached one of the 10 largest modeled liquidation-map clusters (size over min_usd; no symbol = BTC, ETH, SOL), sent 3 min later with modeled vs actually liquidated USD. Returns the signing secret ONCE: verify header X-LiquidVision-Signature = hex HMAC-SHA256(secret, raw body). Needs a Pro or Bot key.

[account] Remaining prepaid credits on a key bought with buy_credits. Free/Pro/Bot keys are metered by daily quota instead (X-Raw-Quota-* headers) and return a note saying so.

Permanently delete one webhook alert subscription. DESTRUCTIVE and not reversible: deliveries to its URL stop at once and its signing secret is discarded (create_alert issues a new one). Only a subscription owned by the caller's API key can be deleted; an id belonging to another key returns {"deleted": false} and changes nothing. id: the number returned by create_alert or list_alerts. Needs a Pro or Bot key (X-API-Key header of the MCP connection, or api_key).

AI risk-intelligence answer: ask e.g. 'Why did BTC drop 4%?' and get a concise grounded explanation (liquidation cascades, funding extremes, where the positions sit, next risk zone) computed from live derivatives data across 13 venues (CEXs, Hyperliquid and GMX).

[measured] The public accuracy audit: our open interest and funding vs Binance, Bybit, OKX and Hyperliquid's own endpoints, sampled every 10 min (diff %, snapshot age, 24h mean/max), plus how the modeled liquidation map scored against realized liquidations. Read this before trusting any number here with money.

Get a free LiquidVision API key for this agent's owner, by email, without a browser. Works for a NEW email only (an existing account's key is never returned; its owner sees it at liquidvision.app/data). Use the key as X-API-Key on REST or on this MCP endpoint: 120 req/min instead of the shared anonymous budget, plus 100 metered calls a day (history, positions: get_funding_history, get_open_interest_history, get_funding_settlements, get_hl_whales, get_history...), and to upgrade later to Pro/Bot (10,000 / 100,000 metered calls a day, exports, webhook alerts). Ask the user for the email first; do not invent one.

Arbitrum perps board: GMX v2 open interest per coin and its share of every tracked venue, GMX funding next to the CEX median (8h basis), on-chain liquidations over 24h. Ask this for 'how is Arbitrum positioned vs CEXs'.

Derivatives cash flow per exchange: net USD into/out of open interest over 1h/4h/24h windows, per core symbol. Positive = money flowing in.

Cumulative volume delta split by trade size (retail <$10K, mid <$100K, large <$1M, whale >$1M). Whale-vs-retail divergence is the usual smart-money reading; buckets are per published row and only Binance aggregates taker orders, so compare buckets within one exchange (exchange=binance) rather than across venues.

[measured] Detective: what the watched large holders did. Private wallets with a large share of a token's supply (Ethereum, Solana) and Hyperliquid accounts with a large share of a contract's open interest are watched live. Event types: to_exchange (a watched wallet sent tokens to a recognised exchange, custodian, market maker, DEX pool or bridge: a possible sale, reported at any size), large_move (at least $100k to an address we do not recognise), hl_position (a watched Hyperliquid position opened, closed or changed by a quarter and $1M+). Each event carries the sender's share of supply when it was added to the list, the destination label and its source, amount, best-effort USD and an explorer link. Filters: hours (1-720), chain (ethereum|solana|hyperliquid), type, symbol (coin). The watch list itself: GET https://liquidvision.app/api/v1/detective/watch.

[measured] Footprint chart data: for each candle, taker BUY and taker SELL volume in USD and the trade count at each of 30 price levels, from the archived trade streams of Binance, Bybit, OKX, Bitget and Gate. Use it to see at which prices aggressive buyers or sellers were absorbed; for the net flow over time by trade size use get_cvd. symbol: BTCUSDT, ETHUSDT or SOLUSDT only. minutes: look-back window, 15 to 10080. step: candle size in minutes, 1 to 240.

[measured] Funding carry screener on REALIZED numbers, out of sample: for each coin, the venue pair a desk would have put on a week ago (short where funding WAS richest, long where it was cheapest, chosen on days 14-7 ago: selection_week_apr) and what that pair then earned from funding over the last 7 days (realized_7d_*, negative when the spread reversed) and 30 days, annualized; positive_share_7d = share of the week's hours the pair was net positive (0.5 is a coin flip); current_spread_apr = the same pair right now; mark_gap_bps = short-venue mark over long-venue mark (positive = entry in your favour). Use get_funding_dispersion for the live spread across all venues. Everything is gross: before taker fees, basis moves and two margin accounts. Not financial advice.

Coins ranked by how far funding disagrees ACROSS exchanges.

For each coin: the spread in APR points between the venue charging the most
and the one paying the most, plus which venue to hold the long on and which
to hold the short on. Every leg is annualized with its own settlement cycle
(1h / 4h / 8h), and the ranking uses the median rate over `window` minutes
rather than the latest print, which is noisy for hourly contracts.

[measured] Funding rate over time for one contract, one row per venue per time bucket: ts (ms), exchange, funding_rate (per settlement interval, a fraction), mark_price and funding_hours (that contract's interval: 1, 4 or 8). Annualize as rate * 8760 / funding_hours. This is the LIVE predicted rate sampled over time; for the rates actually paid at each settlement (2 years back) use get_funding_settlements. symbol: BASEUSDT. minutes: 5 to 20160.

Latest perpetual funding rates and mark prices for every tracked exchange/symbol pair. Positive rate = longs pay shorts.

Settled funding rates, one row per settlement, from Binance and Bybit history — up to 730 days for any pool coin (e.g. BTCUSDT). The backtest series; get_funding_history is the live 10-second archive since July 2026.

[measured] The HIP-3 board: Hyperliquid's builder-deployed perp markets (equity indices, stocks, commodities, pre-IPO names; coins named dex:COIN such as xyz:SP500) with what the live list alone cannot tell you. Per dex: OI, share, 24h/7d OI change, 24h volume, liquidations. Per market: mark, 24h change, OI and its 24h/7d change, 24h volume and turnover (volume/OI), hourly funding and APR, liquidations 24h/7d (count, USD, long share), and gap_bps = HIP-3 mark vs the mean mark of other venues listing the same underlying (Lighter, Ostium, gTrade). leaders lists OI growth/decline, funding extremes, most liquidated, widest gap. dex filters to one builder (xyz, io, para...); sort = oi | volume | funding | liq | gap | oi_change; limit caps the markets list. OI changes are null until the archive is that old (started 2026-10-31). No key needed.

[measured] Historical rows as CSV, for backtests. dataset: liquidations | funding_settlements (since 2024-09) | funding_hourly | open_interest_hourly | hl_positions (Hyperliquid whale positions with liquidation prices). start/end are ISO dates (UTC). Without a key or on the free tier only the last 24 hours are available; Pro: any 31-day window; Bot: full history. The key is read from the MCP connection's X-API-Key header, or pass api_key. limit caps rows here (max 5000); for bulk use download_url with the same key.

[measured] One Hyperliquid account by address: live positions (size, entry, mark, liquidation price, distance to liquidation, leverage, unrealized PnL), account value, and its archived position snapshots. History depth: 24 h free, 31 days Pro, everything since 2026-09-27 on Bot (key from the MCP connection's X-API-Key header, or api_key).

[measured] Perp markets deployed on Hyperliquid by third parties (HIP-3 builder dexs): equity indices, single stocks, commodities and pre-IPO names such as xyz:SP500 or xyz:GOLD. For each dex: its markets (coin named dex:COIN), open interest in USD (one side), 24h volume, mark price, hourly funding and its APR. These are NOT in get_open_interest or get_funding_rates, which cover Hyperliquid's own crypto listing. Live from Hyperliquid, cached 2 minutes. No arguments.

[measured, RISK AHEAD, Hyperliquid] REAL open positions on Hyperliquid binned by the liquidation price the exchange itself reports: the 10,000 largest accounts plus every account that traded since the reader started (~10,000 accounts, ~40-45% of venue open interest, stated in covered_long_pct/covered_short_pct). Not a model. Use this for Hyperliquid; use get_lighter_liquidation_map for Lighter and get_liquidation_map (modeled) for the CEXs. Returns long/short notional per bin, cumulative curves and the largest positions. range: 12h | 1d | 3d | 7d | max (±100%, recommended: large accounts run low leverage).

[measured] Leveraged positions of the 10,000 largest Hyperliquid accounts with the exchange-reported liquidation price, closest to liquidation first (sort='value' = largest first). Filters: coin (BTC), side (long|short), min_usd, max_dist_pct. Each row carries the account address; pass it to get_hl_account. Snapshot age is in age_s.

[measured, RISK AHEAD, Lighter] REAL open positions on Lighter (zk perp DEX; crypto plus equity indices, stocks, forex, commodities) binned by Lighter's own liquidation price. Not a model. Use this for Lighter; get_hl_liquidation_map for Hyperliquid; get_liquidation_map (modeled) for the CEXs. Accounts are discovered from the public trade stream, so coverage (covered_long_pct/covered_short_pct of venue open interest) grows with time. Returns long/short notional per bin, cumulative curves and the largest positions. range: 12h | 1d | 3d | 7d | max.

[measured] Leveraged positions on Lighter with the exchange-reported liquidation price, closest to liquidation first (sort='value' = largest first). Filters: coin (BTC), side (long|short), min_usd, max_dist_pct. address is the Lighter account index, l1_address the wallet. Snapshot age is in age_s.

[measured, PAST] Liquidation CASCADES as discrete events: bursts of liquidations detected against each coin's own 7-day normal minute. Use it for "when did forced selling snowball and what did price do"; not for totals (get_liquidations_summary), raw orders (get_liquidations) or open positions at risk (*_liquidation_map). Each event: coin, which side broke, start, how long it ran, total and peak-minute USD, multiple of the coin's normal minute, the mark-price move during it and in the hour after. market_wide lists minutes when three or more unrelated coins cascaded together. hours: display window (baseline is always the last 7 days).

[modeled, RISK AHEAD, CEX, over time] The same estimated CEX map as get_liquidation_map but as a time series: grid[t][b] is the estimated USD of leveraged positions with their liquidation price in price bin b at time cols[t] (ms), bins[b] the bin's centre price (240 bins fitted to the traded range), price[t] the close. A level disappears once price trades through it. Call this only to see WHEN a cluster formed and whether it survived; for the current picture call get_liquidation_map (far smaller answer, split by leverage and exchange). The answer is large (about 35,000 numbers). symbol: BASEUSDT. exchange: all | binance | bybit | okx. range: 12h | 1d | 3d | 7d.

[modeled, RISK AHEAD, CEX, now] ESTIMATED open leveraged positions on Binance/Bybit/OKX per price bin at their liquidation price, from candles and open-interest deltas (the CEXs do not publish positions). Current snapshot only; for its evolution over time use get_liquidation_heatmap; for REAL positions use get_hl_liquidation_map (Hyperliquid) or get_lighter_liquidation_map (Lighter); for what already got liquidated use the PAST tools. Returns notional per bin split by leverage tier (5/10/20/25/50/100x) and by exchange, plus cumulative long/short curves from the current price outward. exchange: all | binance | bybit | okx. range: 12h | 1d | 3d | 7d. It ranks levels ("if price reaches this level, how much is there": top quarter of traded levels holds ~30-32% of liquidated USD, chance ~24%); it does NOT predict where price will go. Live scoring: get_accuracy.

[measured, PAST] Individual liquidation ORDERS that already happened (side, price, qty), newest first, at most limit rows (max 2000). This is the raw feed: use it when you need the orders themselves. For totals use get_liquidations_summary; for cascades as events use get_liquidation_cascades; for where OPEN positions would be liquidated use a *_liquidation_map tool. Market-wide, 500 rows is only a few minutes of a busy hour: pass symbol (BTCUSDT) and/or min_usd to look further back. side=long means a long position was liquidated.

[measured, PAST] TOTALS of liquidations that already happened: rekt (USD liquidated in the last 1h/4h/12h/24h, split long/short), by_symbol and by_exchange over the last minutes, and series (30-minute buckets). The one to call for "how much was liquidated"; it does not return individual orders (get_liquidations) or events (get_liquidation_cascades), and says nothing about open positions (the *_liquidation_map tools). Totals are a floor: Binance and Bitget throttle their feeds. side=long means longs were liquidated. minutes: 1 to 1440.

Latest global long/short account ratio per exchange/symbol.

Deterministic market brief for a perp symbol: 24h liquidations (totals, by exchange), current funding + 24h extremes per venue, OI by venue with 1h/4h/24h changes, net position flows, modeled liquidation risk zones above/below price, and top-trader positioning. Free, no LLM involved — ideal input for your own reasoning.

Cross-venue order book microstructure for BTC, ETH and SOL.

Per venue: spread in bps, immediately-executable depth in USD and the bid
share of it, measured over a band narrow enough that every venue's stream
captured it, each venue around its own mid. Also the basis - how far apart
venues price the same asset - plus a per-minute spread history.

Does NOT answer "what moves price 0.5%": the streams only carry the
innermost levels, so that would be extrapolation, not measurement.

Per-venue open-interest ranking for a perp symbol (e.g. BTCUSDT): OI in coins and USD, market share %, 1h/4h/24h OI change % and USD flow, and OI / 24h-volume ratio per exchange, plus market totals.

Latest open interest (contracts and USD) per exchange/symbol.

[measured] Open interest over time for one contract, one row per venue per time bucket: ts (ms), exchange, oi (coins, one side), oi_usd. Use it to see how positioning built up or unwound; for the current snapshot with shares and 1h/4h/24h changes use get_oi_board, for every contract at once get_open_interest. symbol: BASEUSDT (BTCUSDT). minutes: look-back window, 5 to 20160 (14 days); snapped to a fixed set of windows.

[measured] Resting limit orders over time on ONE exchange: cells of (time bucket, price bucket, quantity in coins) from archived order-book snapshots, with price_min, price_max, price_step and time_step (seconds) describing the grid. Shows where bid and ask walls stood and when they were pulled; not liquidations (see get_liquidation_heatmap for those) and not trades (get_footprint). exchange: binance | bybit | okx. symbol: BTCUSDT, ETHUSDT or SOLUSDT only. minutes: 1 to 1440.

Per-coin positioning regime from price direction crossed with open interest.

price up + OI up = new longs; price up + OI down = short covering (a rally with no new buyers); price down + OI up = new shorts; price down + OI down = long flush. Returns a market summary (coins and net OI per regime) plus per coin the price change, OI change and current OI over the window.

[measured] When did the market look like this before, and what did price do next? Scans our hourly archive for hours whose positioning resembled now (OI change 24h within 3 pp, funding APR within 15 points and same sign, long share within 5 pp, liquidation volume within x2.5), one per day, and returns the 24h and 72h price outcome of those look-alikes: count, share that went up, median and quartiles, worst and best, next to the unconditional base rate over the same archive. verdict says whether the sample is big enough to mean anything. Frequencies from our archive (since 2026-09), NOT a forecast. Metered: free key (100/day), Pro/Bot, or x402.

Order-book wall lifecycle heuristic: large levels classified as active_wall / absorbed / pulled_wall / suspected_spoof based on whether trades executed into them before they vanished.

Crypto-related equities (COIN, MSTR, miners) and spot BTC/ETH ETF quotes: price, day change %, range, volume. ~15-min delayed.

[measured] Tokenized stocks, indices, ETFs, commodities, FX, rates and pre-IPO names traded as PERPS on DEXs (Lighter and Hyperliquid builder markets): one row per underlying (NVDA, TSLA, S&P 500, Gold, EUR/USD, Anthropic...) with every venue's mark, open interest, 24 h volume, funding APR and the price gap between venues in bps. asset_class filters: equity | crypto_equity | etf | index | commodity | fx | rates | preipo. Use get_hl_builder_markets for the raw Hyperliquid HIP-3 list by dex.

Upcoming token unlock events for pool coins, nearest first: next unlock timestamp and token amount per public vesting schedule.

Top lead traders (OKX copy trading, public data): total PnL, win rate, AUM, copiers, live open positions (pair, side, leverage, entry, uPnL) and recently closed trades with realized PnL.

The webhook alert subscriptions that belong to the caller's API key (read-only): id, type, url, filters (symbol, min_usd, x_normal, max_dist_pct) and delivery health: last_status (HTTP code or error of the last delivery), fails (consecutive failures; a subscription pauses itself at 20) and active. Call it to find the id for delete_alert or to check why alerts stopped arriving. Needs a Pro or Bot key, from the MCP connection's X-API-Key header or the api_key argument; without one it returns an error object, not an exception.

[measured] Monthly Parquet history packages for backtests: liquidations (11 venues), funding hourly and settlements, open interest hourly, Hyperliquid and Lighter positions with liquidation prices — one file per dataset per month since 2026-07 with rows, bytes and sha256. Each entry has url (download with the API key in X-API-Key) and available for the caller's tier: Pro = this month and the previous one, Bot = all, free = none (use get_history for the last 24 h).

[measured] What exists: every venue we track and, for each, the contracts with live open interest right now (BASEUSDT symbols), plus the top-100 coin pool. Call this first when unsure which symbol or exchange string another tool accepts.

[measured] One rule tested on up to 100 coins in a single call: on which coins it was true in our hourly archive, what price did over the next 24 h and 72 h per coin (share up, median, worst, best) next to each coin's base rate, the POOLED outcome across coins with the number of distinct days behind it (coins move together, so days, not episodes, are the sample), and true_now: the coins the rule is true on right now. Use it to find where a setup has occurred and where it is live, instead of calling test_rule per coin. Frequencies from our archive, NOT a forecast; the note explains why the best of many scans is not an edge. Priced: $1.00 per call (x402, USDC on Base or Arbitrum), 200 prepaid credits (one $1 pack), or a Pro/Bot key. Takes 10-40 s on a cold cache.

[measured] Test a trigger against our hourly archive BEFORE acting on it or alerting on it: when was this rule true before, and what did price do next? Returns matched hours, independent episodes (one per 24 h) with each one's 24 h and 72 h return, the outcome statistics (share up, median, quartiles, worst, best), the unconditional base rate over the same hours, whether the rule is true right now, and a verdict on sample size. The rule syntax is the condition-alert syntax, so a rule that tests well becomes create_alert(type= 'condition', ...) with the same metric/op/threshold. Frequencies from our archive (hourly, since 2026-07 for BTC/ETH/SOL), NOT a forecast. Priced: $0.10 per call (x402), 20 prepaid credits, a free key's daily quota (100 metered calls), or Pro/Bot.