Changelog

​

Candles: 5s and 30s intervals

• candleSnapshot (POST /info) and the candle / allCandles WebSocket channels accept two new intervals, 5s and 30s, on top of Hyperliquid’s. Same candle shape; 5s and 30s WebSocket buckets are final at their closed frame (reconciled: false).
• Open and close now follow the exact execution order of trades inside a block, as on Hyperliquid, for candles from 1 October 2026 onwards.

​

historicalOrders: open and triggered states of trigger orders

• Stops and take-profits now come with their open and triggered states next to their final state, as in Hyperliquid’s response. These states carry triggerPx, triggerCondition and the creation timestamp, which the final filled state of an executed stop resets (triggerPx: "0.0", execution timestamp). A trigger order spans several rows; limit counts orders.
• In window mode, an order is returned when any of its states falls in the window, so a window around a stop’s placement now returns its open state.

​

Fills: TWAP parent and builder fee everywhere

• fills and recentFills (POST /info, compact shape) now return twapId on every row (null = ordinary fill), plus builderFee and builder. Fills of a TWAP carry their parent as soon as they are indexed.
• userTwapSliceFills is served from the indexed fills again: it had been reading a table that stopped being fed on 2025-10-11 and returned nothing for recent TWAPs. New optional twapId parameter to get the slices of one parent; startTime / endTime are inclusive bounds on the fill time.

​

Orders: complete traversal of a time window

• historicalOrders: new startTime / endTime (inclusive, on the status timestamp) and cursor. In this window mode the result is sorted by oid descending and paged with the X-Has-More / X-Next-Cursor response headers, so a wallet with more than limit orders in the window can be read completely. The status returned is the latest one within the window. Without these parameters the behaviour is unchanged.

​

Pagination headers on POST /info

• When an indexed type is paginated, the response now carries X-Has-More, X-Next-Cursor and, where available, X-Total-Count headers. The body keeps its shape. Applies to fills, recentFills and historicalOrders in window mode.

​

Leaderboard across every address

• GET /users/leaderboard: new window=7d|30d|all mode. The ranking is pre-computed over every address that traded in the window and paginated with page / limit (up to 500 per page, total_count and has_more in the envelope). New metrics win_rate, pnl_pct (net PnL / traded volume) and unrealized_pnl, next to volume, pnl and trades; each row also carries fills, realized PnL, fees, wins, losses, account value and refreshed_at. 7d and 30d are refreshed every 5 minutes, all every hour. min_trades (default 10) keeps one-trade wallets out of the win_rate / pnl_pct rankings. Without window the endpoint behaves as before (live top 100 over the last hours hours).

​

New: dataCoverage

• dataCoverage (POST /info) returns, per dataset, the start of complete coverage and the latest event timestamp present in the region answering, with the delay in seconds: the way to tell an empty answer from a window that is not covered.

​

Per-user funding, rebuilt

• userFunding / accountFunding (POST /info) and GET /funding/userFunding: every payment now comes from Hyperliquid’s official node event archive. time is the exact settlement timestamp in milliseconds; delta gains coin, fundingRate and nSamples (always null); hash is the zero hash, as on Hyperliquid. Top-level coin, usdc and szi are unchanged. Coverage starts 2025-09-27 10:00 UTC; an earlier window returns []. Before this update, payments earlier than 3 September 2026 could carry an inverted sign, a position size taken at the wrong moment, or blank rows.

​

Native ledger

• userNonFundingLedgerUpdates (POST /info) now serves the core Hyperliquid ledger in the native {time, hash, delta} shape: deposits, withdrawals, sends, spot, internal, sub-account and perp/spot class transfers, vault create/deposit/withdraw/distribution/commission, liquidations, borrow/lend, staking transfers, rewards and gas auctions. One entry per event and per address involved. Coverage starts 2025-09-27 09:29 UTC. EVM-side transfers stay on /evm/ledger/*.

​

New WebSocket channels

• userFundings: snapshot of the 100 most recent payments, then one frame per hourly settlement, about 30 s after the hour.
• userNonFundingLedgerUpdates: snapshot of the 100 most recent ledger entries, then new entries within seconds.
• userEvents now also carries funding settlements ({"funding": {...}}), one frame per payment.
• userFills: when the account has no recent live fill, the snapshot is served from storage instead of being empty.
• candle / allCandles: new 30m interval.

​

Fills

• crossed is correct on every fill stored since 2026-09-24 15:53 UTC (earlier stored fills read false; a historical correction is scheduled).
• twapId is always present in Hyperliquid-shaped fills, null outside a TWAP. cloid is never "<nil>": omitted, or null in the compact shape.
• time is the exact millisecond on every fills endpoint, with a stable ISO 8601 format on repeated queries.
• Compact fills / recentFills rows gain crossed (0/1) and cloid.

​

Orders

• historicalOrders: lookbackDays is optional (1–365). Without it, the limit most recent orders are returned whatever their age (default 2000, max 5000), as on Hyperliquid. Numbers use Hyperliquid’s format ("93604.0"); cloid and tif are always present, null when absent.

​

Market history

• Seven new endpoints under /market: coins, bbo/{coin}, spread/{coin}, depth/{coin}, slippage/{coin}, book/{coin}, liquidity-report/{coin}. Coverage April 2024 to February 2026 from the market data archive, and continuously since 23 September 2026 from the full order book of our own node for the book endpoints. See the Indexed Data REST reference.

​

Borrow/lend reserve history

• borrowLendReserveHistory (POST /info): borrow and supply rates, balance, utilisation, oracle price, LTV, total supplied and borrowed per reserve, one sample per minute since 23 September 2026. Filter by token, resample with interval (1m to 1d, last sample of each bucket), window with startTime / endTime, up to 10,000 rows.

• liquidations WebSocket: new users parameter, up to 50 addresses, matching the liquidated user or a liquidator. Combines with user, amount_dollars and builder.

​

New endpoint

• POST /users/batch/performance: Same stats as GET /users/{user}/performance for up to 500 wallets in one call: one entry per submitted address, in submission order, computed in a single query (~300 ms for 500 wallets). window = 7d, 30d, 90d, all. include_drawdown: true adds the realized-PnL max_drawdown per wallet (heavier). The equity-based drawdown stays on the per-wallet endpoint. Billed 3 credits per wallet submitted, the same as one per-wallet call per address (also applies to POST /builders/{builder_address}/external-activity). See the Social Trading guide.

​

WebSocket candle / allCandles: final frame per bucket

• Every bucket now ends with one explicit final update carrying closed: true, sent ~2 s after T, whose o/h/l/c/v/n are recomputed from the stored trades (reconciled: true). A bucket is final when you receive that frame: see candle and allCandles.
• Late or out-of-order trades are now applied to their own bucket instead of resetting the current one, and a batch of trades spanning two buckets emits an update for both. This fixes closed bars that could end short of the true trade count without any signal.
• reconciled: false on a final frame means storage was temporarily unavailable and the frame reflects the live state only; a later closed frame for the same t supersedes it.

​

Builders: external activity

• POST /builders/{builder_address}/external-activity, Send your builder address and up to 500 wallet addresses; for each wallet you get the fills attributed to other builder codes (externalBuilders[], with fill count, notional, coins, first and last fill) and, unless includeNoBuilder=false, the fills carrying no builder code at all (noBuilder), the official Hyperliquid UI, bots, any tool outside the builder program. own reports the wallet’s activity under your code for context, and flagged users come first, sorted by external volume. summary counts the wallets in each situation. timeframe = 1h, 24h, 7d (default), 30d. Builder attribution starts 2025-10-10. See the Builders reference.

​

Social Trading endpoints

​

New endpoints

• GET /users/{user}/equity-history: Account value over time, one point per hourly snapshot (interval=1h) or per day (interval=1d). Each point carries account_value, cash, position_value, unrealized_pnl, account_leverage and n_positions; summary adds the change over the window and its max drawdown. window = 7d (default), 30d, 90d, all.
• GET /users/{user}/pnl-calendar: One row per trading day: realized_pnl (from fills), fees, funding, net_pnl = realized − fees + funding, cumulative_net_pnl, fills, volume and equity_close. totals counts profitable and losing days. Funding follows the Hyperliquid sign convention (positive = received). window = 7d, 30d (default), 90d, all.
• GET /users/{user}/account: Latest known state: account_value, cash, position_value, unrealized_pnl, account_leverage (open notional ÷ account value) and every open position with leverage, leverage_type, entry_px, mark_px, unrealized_pnl, roe and accrued funding. as_of is the snapshot time.
• GET /users/{user}/top-trades: Best round-trip trades with coin, direction, leverage, pnl, notional, entry/exit price and closed_at. Open positions are ranked alongside by unrealized PnL (status: "open", no closed_at) unless include_open=false. sort = pnl (default), pnl_abs, notional, loss; limit 1–50 (default 5).

​

Updated endpoint

• GET /users/{user}/performance: New window parameter (7d default, 30d, 90d, all); start_time / end_time still take precedence when supplied. Response gains volume, longs / shorts / long_short_ratio / long_pct, total_fees, total_funding, best_trade_pnl / worst_trade_pnl, and a second drawdown measured on real account value (equity_max_drawdown_usd, equity_max_drawdown_pct, equity_snapshots) alongside the existing max_drawdown on the realized-PnL curve. Existing fields are unchanged. An invalid address now returns 400 instead of 500.

​

Coverage

• Round-trip trades (/performance, /top-trades): from 2025-09-10.
• Account snapshots (/equity-history, /account, equity_close, per-trade leverage): from 2026-09-05, hourly, accumulating from there. Leverage on trades closed before 2026-09-06 is null.
• account_value covers the perp account only: spot balances and HIP-3 builder-dex positions are not included yet. A trader with no position and no perp balance returns account_value: 0 with perp_account_empty: true, which is a real zero rather than missing data.

​

HIP-4 permissionless support

​

New endpoints

• GET /hip4/providers: Per-provider trading statistics aggregated from fills (markets_traded, fills, volume_usdc, unique_users, fees, last_trade), sorted by volume_usdc descending. Filters: venue (a venue name, or oracle), start / end (ISO datetime, inclusive), limit / offset. Settlement rows are excluded from all aggregates; a new permissionless venue appears automatically after its first fill.
• GET /hip4/deployers: The permissionless deployer registry, refreshed from chain metadata every ~10 s: deployer, venue, fee_scale, sub_deployers (a JSON string to parse: [action, [addresses]] pairs of delegated permissions per venue), updated_at.

​

Updated endpoints

• GET /hip4/markets (alias /hip4/outcomes): Three new fields on every market row: venue ("" for oracle-emitted markets), deployer ("" when venue is empty), deployer_fee_scale (0 when not applicable). Everything else unchanged.
• GET /hip4/fills: Two new fields on every fill row: venue and deployer, joined from the market of the fill’s outcome. The existing market_name / market_description fields are unchanged.
• GET /hip4/questions: Two new fields on every question row: venue and deployer, derived from the question’s outcomes (resolved via the fallback outcome’s market). Legacy questions report venue="oracle", deployer="".

​

Order type on fills (2026-08-28)

• orderType (string or null): the originating order’s type, resolved server-side from the order status by (user, oid). Observed values include Limit, Market, Stop Market, Take Profit Limit (an open set, not an enum).
• With the default false, the response is byte-identical to before, the key is absent, not null, so existing integrations are unaffected.
• orderType is null when the originating order status is unknown: orders placed before 2026-06-27 (start of order-status coverage), plus a small share of gaps in that dataset.
• Cost: roughly tens of milliseconds per page of 1000 fills.

• agentSendAsset: agent transfers between dexes for the same user.
• hip3LiquidatorTransfer: deposit/withdraw of principal on the backstop of a HIP-3 dex.

​

New endpoints

• GET /evm/hip3/backstop/transfers: Filterable list of backstop principal transfers (filters: dex, signer, is_deposit, time range).
• GET /evm/hip3/backstop/transfers-summary: Per-dex aggregate: total_deposited_usdc, total_withdrawn_usdc, net_principal_usdc, unique_signers.

• GET /evm/hip3/backstop/health: Health overview of every observed HIP-3 backstop. Returns one row per dex (combined principal + fill counters, last/first fill, fees paid, active coins).
• GET /evm/hip3/backstop/{dex}/health: Same payload for a single dex. Returns 404 if no backstop activity has ever been observed for that dex.
• GET /evm/hip3/backstop/{dex}/fills: Paginated stream of raw fills absorbed by the backstop. Filters: coin, side, time range.

​

Updated endpoints

• GET /evm/ledger/transfers: Response extended with source_dex and destination_dex (populated for agentSendAsset). action_type=agentSendAsset is now a supported filter value.
• GET /evm/user/{address}/ledger-events: New agent_send event type, exposing source_dex and destination_dex.