type parameter. GET /v1/funding/rates returns the authoritative venue set: every distinct exchange value in that response is queryable.
The funding book is not crypto-only. Tokenized equity, commodity, index and FX
perps trade on the same venues and are returned by default. Every row carries
an
asset_class; pass asset_class=crypto to exclude the rest.rate is a fraction, not a percent, and it is charged once per
interval_hours. Annualize with rate × (24 / interval_hours) × 365.
Intervals differ per contract. The live book contains 1h, 2h, 4h, 8h and 24h
contracts, and the same venue runs different intervals on different symbols.
Never assume 8 hours.coin names an asset (since 2026-09-27). It used to match the base_coin
string, so coin=PEPE missed every PEPE contract a venue lists as
1000PEPE or KPEPE; it now returns all of them, and coin=1000PEPE
returns the same set. Rows carry asset_id, asset_symbol,
lot_multiplier, instrument_status and is_live from Sharpe’s instrument
master, so a client never has to parse a symbol to tell a lot contract or a
halted market apart. The five fields are null for a contract the master
does not know yet. rate is a fraction of notional, so lot contracts compare
with the rest without rescaling.Get funding rates
GET /v1/funding/rates
A single endpoint that returns current snapshots, accumulated sums, or historical time series depending on the type parameter.
Query parameters
The kind of funding rate data to return. One of
current, accumulated, history. Defaults to current.The coin: a ticker (
BTC, ETH, PEPE), a lot or venue spelling (1000PEPE, kPEPE) or the coin’s full name (bitcoin resolves to BTC). Case-insensitive, one coin per request. coin names an asset, not a base_coin string: every contract of the asset comes back whatever its base_coin spelling, so PEPE, 1000PEPE and kpepe all return the PEPE contracts listed as PEPE, 1000PEPE and KPEPE (read lot_multiplier). A ticker two assets share (BB is BounceBit and BlackBerry) returns both; narrow with asset_class. Optional for current and accumulated. Required for history. A value no perpetual has ever settled returns a 400 error naming what was tried, rather than an empty result set that reads as “this coin has no funding data”; a coin delisted from every venue still resolves, so its history stays queryable. GET /v1/funding/coins lists every ticker carried.Lookback window in days for
type=history. Integer from 1 to 1095. Defaults to 30. Ignored for other types.Narrows the response to one venue. Accepts the slug (
gate-io), the display name exactly as returned in the exchange field (Gate.io), or the ccxt id. Every venue Sharpe fetches is addressable. The registry is not frozen, so read the current list from an unfiltered response rather than hardcoding one. An unrecognized value returns a 400 error.Narrows the response to one underlying category. One of
crypto, equity, commodity, fx, index. Unfiltered by default, so the response carries tokenized equity, commodity, index and FX perps alongside crypto. Applies to type=current and type=accumulated; ignored for type=history. An unrecognized value returns a 400 error.Narrows the response to one collateral convention:
linear for quote-margined contracts, inverse for coin-margined ones. Unfiltered by default. Applies to type=current and type=accumulated; ignored for type=history. An unrecognized value returns a 400 error.Maximum number of results per page.
1 to 5000. When provided, the response includes a pagination object with cursor-based paging.Opaque cursor for fetching the next page. Returned in the
pagination.cursor field of paginated responses.For type=current: orders the rows by their open interest before the page is cut. One of
open_interest (ascending), -open_interest (descending). Rows whose open interest is unknown come last either way. Omit it to keep the book’s order (base coin, venue, symbol).For type=current: an integer from
1 to 100. Answers with up to N highest rows (rate_8h ≥ 0) and up to N lowest rows (rate_8h < 0) instead of the rows (see “Aggregates” below). The other filters narrow the rows first. Cannot be combined with limit, cursor or sort.For type=current, with
extremes only: leaves out rows whose open interest in USD is below this floor. A row with unknown open interest fails any floor above 0. Passing it without extremes returns a 400 error.For type=current: set to
1 to answer with the whole book’s funding (market), and how many tracked venues are live, instead of the rows. Cannot be combined with coin, exchange, margin, asset_class, limit, cursor or sort.For type=current or type=accumulated, with
coin: set to 1 to answer with the coin’s funding across venues (summary) instead of the rows: on type=current its average rate, open interest and next settlement, on type=accumulated its medians and carry spread per window. Cannot be combined with limit, cursor or sort.type=current
Returns the latest funding rate snapshot for every exchange-symbol pair. Stale entries older than 6 hours are automatically excluded. Every scope is answered from the book the funding job writes every 5 minutes.meta.as_of is the newest write among live contracts in the scope, and X-Data-Stale judges it against the 10-minute SLA, so one halted venue’s frozen rows never mark the rest of the book stale (each row still carries its own is_stale). A scope that matches no contract is dated by the book; a scope of only halted or delisted markets has no as_of and says so in meta.warnings. When the database read fails, the answer comes from the last written book at any age, flagged X-Data-Fallback: snapshot and never cached. With nothing to serve the endpoint returns 503 (service_unavailable, Retry-After: 30). A repeated keyed request is answered from a 100-second memo while the book is fresh.
Response fields
The venue’s display name, in its exact stored casing:
Binance, Bybit, OKX, Gate.io, Crypto.com, WhiteBIT, edgeX, tradeXYZ. Match on these values, not lowercase slugs.The base ticker as stored for this contract (e.g.,
BTC, ETH). Lot spellings are kept (1000PEPE, KPEPE), so one asset can appear under several; group contracts by asset_id.The funding rate as a fraction charged once per
interval_hours (e.g., 0.0001 = 0.01% per interval). Not a percentage and not annualized. What it measures depends on the venue, named by rate_kind: on most venues it is the in-progress estimate for the settlement at next_funding_time, which moves until then and can differ from what the venue last charged by a multiple or in sign; on Kraken, Crypto.com and Bitfinex it is the rate charged at the last settlement; on Deribit it is the funding realised over the trailing 8 hours. Settled rates are served by type=history.The venue’s own forecast for the coming settlement, published beside
rate only by Kraken, Crypto.com (an estimate for the next 4-hour interval, not the next hourly settlement), Pacifica and Hyperliquid (where it equals rate). null everywhere else, including Binance, Bybit and OKX.Hours between funding settlements for this specific contract. The live book contains
1, 2, 4, 8 and 24; a single venue can run several intervals across its symbols, so read this per row rather than assuming 8.ISO 8601 timestamp of the next scheduled funding settlement.
null on venues that do not publish one.Underlying category:
crypto, equity, commodity, fx or index. Tokenized-equity perps such as Bybit TSLL are equity, not crypto.Latest one-sided USD open interest (each contract counted once) of this contract, matched on
(exchange, symbol): two contracts of one coin on one venue carry their own figures. null when no open-interest row matched: a minority of rows, mostly on the venues Sharpe captures no open interest for. Never 0 as a stand-in for missing.Market-cap rank of the base coin, where
1 is the largest. null when the coin is unranked. Sort unranked rows last rather than treating null as a rank.Age of the row when the response was generated.
null means the source timestamp was invalid or unavailable.Operational stale threshold for current rates. The writer runs every 5 minutes and allows a 5-minute grace window, so this is currently
600 seconds.true when the row is older than the published freshness target. Rows up to the backwards-compatible six-hour cutoff can therefore be returned while explicitly marked stale.The asset this contract belongs to in Sharpe’s instrument master: the CoinGecko id when the asset was resolved strictly (
bitcoin, pepe), otherwise <class>:<ticker> (equity:bb for BlackBerry). Every contract of one asset carries the same id whatever its base_coin spelling, so group or join on this, never on base_coin. null when the master does not know the contract yet.The asset’s ticker as a trader reads it:
PEPE on the 1000PEPE and KPEPE contracts too. null when the master does not know the contract.Units of the asset per unit the contract quotes:
1000 for Binance 1000PEPEUSDT (its price and size are per 1,000 PEPE); 1 for Gate.io PEPE_USDT. Taken from the venue or a checked alias, never parsed from the symbol. rate needs no scaling. null when unknown.The venue’s own status for the contract as Sharpe’s instruments sync last read it:
trading, halted, delisted, settling or pre_launch. null when the master does not know the contract.true when the contract is trading and the venue’s contract list was read in the last 30 minutes; false when it was read and the contract is not trading (halted, settling, delisted or not launched); null when that is unknown: no read in 30 minutes, a status value Sharpe could not interpret, or a contract the master does not know. Read null as unknown, never as halted.rate on the 8-hour basis: rate × 8 / interval_hours, so contracts on different intervals compare directly. null when the row has no rate or an interval Sharpe does not recognise; never computed from an assumed 8 hours.rate annualized over the row’s own interval: rate × 8760 / interval_hours, a fraction (0.1 = 10%), signed like rate. null when rate_8h is.What
rate (and so rate_8h and apr) measures on this venue: estimate, the in-progress rate for the settlement at next_funding_time (most venues, Binance, Bybit, OKX and Hyperliquid among them); last_settled, the rate charged at the last settlement, next_funding_time − interval_hours (Kraken, Crypto.com, Bitfinex); trailing_8h, funding realised over the last 8 hours, with next_funding_time on a synthetic 00:00/08:00/16:00 UTC grid (Deribit).Response (type=current)
Aggregates: summary, extremes, stats
Withsummary=1, extremes=N or stats=1 (type=current only), data is an object holding the aggregates asked for instead of the rows. They are computed on the server from the same book, so an agent or a client never re-derives them from 13,000 rows. Several can be asked for at once; each covers the request’s scope. summary=1 also answers on type=accumulated (see “type=accumulated” below).
Response fields
Present with
summary=1 (which needs coin). One entry per instrument and asset class, grouped as the funding grid groups its rows: a coin’s contracts by their asset_id, whatever the venue’s lot spelling, so coin=PEPE answers one PEPE entry over the PEPE, 1000PEPE and KPEPE contracts; a stock, index, commodity or FX perp by its underlying ticker, so MEXC’s KOSTOCK and BingX’s NCSKKO2USD join KO. A ticker two asset classes share (CVX the coin and Chevron) answers one entry per class. Each entry carries base_coin (the entry’s ticker, the coin’s asset_symbol when it has one), asset_class and the fields below.Average funding across venues on the 8-hour basis, over one primary contract per venue with a fresh rate. A
trailing_8h rate (Deribit) is realised funding, not a current rate, and counts as stale. Weighted by each contract’s one-sided open interest when at least two venues, and a majority of them, carry open interest; otherwise the median. null when no venue has a usable rate, never 0.Venues left out because their rate was older than 10 minutes, or was a
trailing_8h rate (Deribit), while another venue was fresh.One-sided open interest in USD over every contract of the entry in the funding book, inverse and lot contracts included, the venues kept out of open-interest totals (MEXC) left out: the funding grid’s OI column. It counts only the funding book’s contracts, every asset class, so it is not the market-wide open interest
/v1/global/overview serves per asset. null when no contract reports open interest.The entry’s next funding settlement on any venue: the soonest published
next_funding_time still ahead among its fresh contracts, or, for a venue that publishes none, the next multiple of the contract’s own interval from 00:00 UTC. null when no fresh contract has either.true when next_settlement_at was stepped from an interval rather than published by the venue. null with next_settlement_at.Present with
extremes=N. highest holds up to N rows with rate_8h ≥ 0, highest first; lowest up to N rows with rate_8h < 0, lowest first (full type=current rows, rate_8h and apr included), among rows that are fresh, have a usable interval and clear min_oi_usd. A list is empty when no such row has that sign: lowest is empty when no eligible rate is below 0, never filled with positive rates. omitted counts the rows each check left out: stale, no_interval, below_floor (open interest below the floor, or unknown while the floor is above 0).Present with
stats=1: the whole book’s funding over one primary contract per venue and asset with a fresh rate, every asset class included. An asset is a row of the funding grid: PEPE, 1000PEPE and KPEPE are one asset, so a venue listing two of its spellings counts once. funding_apr_mean and funding_apr_oi_weighted are the mean and the open-interest-weighted 8-hour rate annualized (× 1,095): the computation behind avg_funding_rate on /v1/market/derivatives-overview, without a key. funding_apr_oi_weighted is null when no contract carries open interest. positive and negative count the contracts whose fresh rate is above or below 0. venue_count is the number of venues Sharpe tracks and live_venues those with at least one fresh contract in the book.Response (type=current&coin=BTC&summary=1&extremes=1)
type=accumulated
Returns pre-computed cumulative funding rate sums and settlement counts across multiple time windows. Useful for calculating net funding cost or income without aggregating history yourself. The response carriesX-Data-As-Of, X-Data-Stale, and X-Freshness-Sla-Seconds from the accumulated pipeline itself. Use each settlements_* field with the contract’s native interval_hours before comparing annualized results; each row’s coverage says, per window, how much of the span the contract was expected to settle in its settlements cover (complete from 90%), whether it was listed inside the window, and whether it stopped trading inside it.
Response fields
The contract’s own settlement interval, joined in from the live contract book.
null when no live contract matches (exchange, symbol). Expected settlements for a window are ceil(window_hours / interval_hours). Compare settlements_* against that, never against a fixed 8-hour assumption.Underlying category, joined in from the live contract book.
null when no live contract matches, rather than being defaulted to crypto.Collateral convention, joined in from the live contract book.
null when no live contract matches.false when the accumulated identity has no live contract behind it, meaning a delisted or renamed pair. Those rows keep their accumulated history but carry null metadata.Per window (
1d, 7d, 30d, 90d, 1y): coverage_ratio (number | null), complete (boolean | null), listed_within_window, halted_within_window (booleans). The span runs from the contract’s first settlement to its halt — the time the instrument master recorded it leaving trading, else its last settlement plus one interval once its row stopped updating — or to now, and counts no more than the window’s covered_hours_*, so a hole inside the span reads as missing. null ratio: the contract was not expected in that window at all. With halted_within_window or listed_within_window true, a complete window’s acc_* sums fewer hours than the window (see covered_hours_*): compare acc_* across contracts only on complete windows where both are false.Served beside
coverage, one per window: the hours of funding the settlements summed in acc_* stand for (the sum of their interval_hours). Fewer hours than the window, or than the contract’s span inside it, means missing settlements, including in the middle of the window. Can exceed the window by up to one interval. null when unknown.With
summary=1 (which needs coin), data is { "summary": [...] } instead of the rows: one entry per instrument and asset class, grouped as on type=current, with base_coin, asset_class and, per window (1d … 1y), acc_median_* (the median acc_* over the venues whose history covers at least 90 % of the window, one contract per venue), acc_apr_median_* (the median of the same venues’ acc_apr_*, not the median sum scaled) and carry_spread_apr_* (the highest acc_apr_* minus the lowest among the coin’s trading contracts that cover the window and hold at least $1M of open interest, never negative; null with fewer than two venues). Each is null, never 0, when nothing qualifies.One per window:
acc_* annualised over the days of history behind it, acc / observed days × 365, as a fraction (0.1095 is 10.95 % a year). The observed days run from the contract’s first settlement in the window to its last plus one interval, capped by covered_hours_* and by the window, so a contract listed two days ago is annualised over two days. null, never 0, when the window holds no settlement or no observed day.The contract’s asset in Sharpe’s instrument master, as on
type=current rows. null when no live contract matches or the master does not know it.Units of the asset per unit the contract quotes, as on
type=current rows. Accumulated sums are fractions of notional and need no scaling.GET /v1/funding/rates?type=accumulated
Response (type=accumulated)
type=history
Returns a time series of individual funding rate payments for a specific coin. Use this for charting, backtesting, or building your own accumulation logic. Results are ordered bysettled_at ascending.
A window is served whether or not the coin has that much history: asking for days=1095 on a contract listed 17 months ago is still a 200. Three response headers state what the body actually covers, derived from the returned rows: X-Requested-Start (the window asked for), X-Coverage-Start (oldest settlement returned, absent when the window returned nothing) and X-Coverage-Complete (false when coverage starts more than seven days after the requested start, or when no rows came back). X-Data-As-Of and meta.as_of are when the funding job last completed a run, which is when the history was last extended, not the newest settlement returned. X-Days-Clamped is always false here because /v1 rejects a days above 1095 with a 400 instead of shortening the window.
The
coin parameter is required when type=history. Omitting it returns a missing_parameter error.Response fields
The contract that settled. One coin spans several contracts per venue at times (a USDT and a USDC perp, or
PEPE next to 1000PEPE lots), so series are per (exchange, symbol). History rows carry no other contract metadata; join type=current rows on (exchange, symbol) for asset_id, lot_multiplier and status.Hours the settlement covered for this contract. It can change over the series when a venue re-tiers a symbol, so accumulate per row rather than multiplying by a single interval.
GET /v1/funding/rates?type=history&coin=BTC
Response (type=history)
Get funding fee settlement
GET /v1/funding/settlement
Returns how many dollars actually changed hands at funding settlement, and which side paid it: fee_usd = open_interest_usd × rate, priced at the one-sided open interest (each contract counted once: the side that pays) recorded at or before each settlement. window=current is the next 24 hours of funding at current rates (forward-looking), priced live and labelled estimated; 1d/3d/7d are realised windows. Every window reports long_paid and short_paid alongside net — a $0 net can conceal $50M flowing each way, so no consumer is given net alone.
The response carries a
totals block scoped to the selected class, in the same shape as a row. The five classes (crypto, equity, commodity, fx, index), and crypto + rwa, each reconcile exactly to all, in every window, for net, long_paid and short_paid.Query parameters
Asset-class scope, the same categories as the funding board. One of
all, crypto, equity, commodity, fx, index, or rwa (every non-crypto class at once: equity, commodity, fx, index). Defaults to all.Narrows to coins settling on one venue. Accepts the slug, the display name as returned in
venues[].exchange, or the ccxt id. An unrecognized value returns a 400 error. Implies the per-venue breakdown internally; the response still omits venues unless expand=venues is also set.Sort field for the selected window, optionally prefixed with a hyphen for descending order. One of
base_coin, net, long_paid, short_paid, open_interest_usd, -base_coin, -net, -long_paid, -short_paid, -open_interest_usd. Rows with no value for the sort field always sort last, regardless of direction.Set to
venues to include each row’s per-venue breakdown. Omitted by default to keep the default payload small.Response fields
One-sided USD open interest summed across whichever of this coin’s venues currently have a priced reading, less any venue left out of totals (see “Venue coverage” below).
null only when none do — never 0 as a stand-in for missing.Keyed by
current, 1d, 3d, 7d. Every window carries net, long_paid and short_paid (see “Sign convention”). current additionally carries estimated: true and rate_source (predicted, current, or mixed when a row’s contracts used both; see “What current means”). The three realised windows additionally carry coverage_ratio, partial, complete, listed_within_window and halted_within_window once the window has any priced history at all (see “Partial windows”).Present only when
expand=venues is set. Each entry is { exchange, open_interest_usd, windows } — the same per-window shape as the row itself, scoped to that one venue — plus excluded_from_totals (the reason) on a venue whose dollars the row and the totals leave out.Sign convention
A positiverate means longs pay shorts, so a positive net is a cost to longs, and a negative net means longs were net paid. long_paid and short_paid are both always non-negative; net = long_paid − short_paid recovers the sign. Both legs are always returned alongside net — a net of 0 can conceal $50M changing hands in each direction, so no consumer is given the net alone.
What current means
window=current has not happened yet — it is the next 24 hours of funding if current rates hold, priced live: fee_usd = open_interest_usd × forward_rate × 24 / interval_hours per contract (so a 1-hour contract counts 24 settlements and an 8-hour contract 3), using the freshest OI reading and the forward-looking rate (the venue’s predicted_rate when it publishes one, otherwise its rate — see rate_source). On Bitfinex and Deribit that rate is the last settled and the trailing 8-hour realised rate respectively (rate_kind on type=current), projected forward. It always carries estimated: true. 1d, 3d and 7d are realised: every settlement that actually occurred inside that window, summed.
Partial windows
Open interest only accrues forward from when Sharpe first started tracking a venue’s OI — no exchange backfills OI history. A venue tracked for the last two days has a complete1d window but an incomplete 7d one until a full week of history exists. coverage_ratio is the fraction of the window actually covered by priced history (1 = complete) — a coin, venue or total row weights its contracts by open interest, and a contract that stopped settling (halted or delisted) is only expected until its last priced settlement plus one interval; partial is true when that ratio is below 0.9. A partial window’s net, long_paid and short_paid are still real sums of whatever settled in that shorter span — building history, not a completed total — so treat partial: true as a caveat on the figure, not a reason to hide it. complete is the opposite of partial. halted_within_window is true when a contract priced in the window stopped settling inside it (its funding row left the live book or went stale), which is why its coverage ends at its last settlement plus one interval. listed_within_window is always null here: the settlement data carries no listing time, so a listing is unknown rather than assumed.
Freshness and caching
settlementMeta.updated_at and meta.as_of are when the settlement windows were last refreshed (every 15 minutes; stale after 30 minutes), not the request time. Responses are cached for up to 5 minutes while fresh and 30 seconds once the settlement data is older than 30 minutes; an incomplete answer (open interest or ranks unreadable) is not cached. A failed read returns 503 (service_unavailable, Retry-After: 30).
Venue coverage
Of the 35 venues in Sharpe’s funding book, 24 run an open-interest fetcher capable of pricing a dollar figure here (Nado, LBank, Bitunix, edgeX, Phemex, Lighter Robinhood, Arcus, Ondo Perps, Antarctic, Polymarket Perps and Kalshi have none, so their contracts carry null open interest), and every one stores open interest the same way: one-sided USD notional, each contract counted once — the side that pays, which is what makesfee_usd = open_interest_usd × rate exact. Venues that publish both sides (Bybit, Gate.io, Bitget, Pacifica, Backpack, ApeX, Extended) are read through their own one-sided field or halved. tradeXYZ, Lighter, Variational and Aster, withheld until 2026-09-27 for reading at half the figure other sources show, contribute again: their figures were one side all along, and those sources publish both. MEXC is the exception: its open interest moves 7.00x the position change its own trade feed reports, so its dollars appear only in the per-venue breakdown (venues[].excluded_from_totals) and are left out of every row’s net, long_paid, short_paid and open_interest_usd and of totals; settlementMeta.oi_excluded_venues names it with the reason. Its funding rates are unaffected and still appear on /v1/funding/rates.
Response
List available coins
GET /v1/funding/coins
Returns every distinct base coin in the current funding-rate book with a display name and a same-origin logo path, ordered by symbol. Use this to populate coin selectors without downloading the full funding book.
Response fields
Response

