KerbTape

Pro and data API

Everything on the Weekend Board and the Whales and Liquidations pages is public JSON and live streams. Build on it, wire it into your own tools, or tell us what your desk needs.

Data API

Read-only, JSON over HTTPS, CORS-safe GETs. Live endpoints are Server-Sent Events; reconnect with Last-Event-ID to resume. Every body carries generated_at, data_as_of and stale, so you can tell a delayed figure from a live one.

Weekend Board

GET/api/board.jsonEvery tracked market: last close, live price, gap, session.
GET/api/m/{coin}.jsonOne market from the board.

Markets (v2)

GET/api/v2/markets.jsonHIP-3 and crypto markets with session state and gaps.
GET/api/v2/listings/new.json?days=7New listings: first trade, OI, volume, funding.
GET/api/v2/carry.jsonSpot↔perp and perp↔perp funding pairs, 7d/30d averages, basis.
GET/api/v2/preipo.jsonPre-IPO and pre-launch markets with published references.

Whales and position events

GET/api/flow/v1/whales.json?window=7dLeaderboard addresses with live positions.
GET/api/flow/v1/events.json?min_usd=100000Position events: open, increase, reduce, close, flip.
SSE/api/flow/v1/events/streamThe same events, live (Server-Sent Events, resume with Last-Event-ID).
GET/api/flow/v1/wallet/{address}.jsonOne address's open positions, with Hyperliquid's liquidation price.
GET/api/flow/v1/wallet/{address}/events.jsonOne address's event timeline.

Liquidations

GET/api/flow/v1/liquidations/tape.jsonInferred liquidations, no addresses.
SSE/api/flow/v1/liquidations/streamLive liquidation tape.
GET/api/flow/v1/liquidations/cascades.jsonCascades: clustered liquidations per coin.
GET/api/flow/v1/liquidations/heatmap.jsonLiquidation levels by price band.

Fees

GET/api/fees.jsonThe published fee schedule: tiers, growth-market rate and its review date, post-only rate, approved maximum, referral terms.
GET/api/fees/{address}.jsonYour fee tier, 30-day routed volume, distance to the next tier and any fee waiver (waived_until).

Historical data (API key)

GET/api/data/v1/datasets.jsonCatalogue: datasets, columns, sources, methods.
GET/api/data/v1/{dataset}.json?from=YYYY-MM-DD&to=YYYY-MM-DDRows as JSON, paged with next_cursor.
GET/api/data/v1/{dataset}.csv?from=…&to=…&cursor=…The same rows as CSV (next page in X-Next-Cursor).
GET/api/pro/exec/report.json?days=30Terminal fills vs arrival: opt-in, anonymised aggregate.

Every public endpoint (102)

Machine-readable: /api/openapi.json (OpenAPI 3.1).

Historical data (v1, API key) · 3
GET/api/data/v1/datasets.jsonHistorical data API catalogue: datasets, columns, sources, methods and coverage per dataset (markets, from, to, rows) (no key needed to read it).
GET/api/data/v1/{dataset}.csvHistorical dataset as CSV (API key required); the next page's cursor is in X-Next-Cursor and Link.
GET/api/data/v1/{dataset}.jsonHistorical dataset as JSON (API key required): weekend_premium, overnight_premium, bounds, reanchors, reopen_gaps, funding_sessions, liq_cascades, scorecard_rows. Query: from, to (YYYY-MM-DD or ISO UTC), coin, limit (≤1000), cursor.
Execution report · 1
GET/api/pro/exec/report.jsonTerminal fills vs arrival: anonymised, opt-in, verified uploads only. Fixed windows of 7 or 30 days ending at the last completed ISO week (Sunday, UTC), so the figures change once a week (STRESS-W5 P1-6). `days` is accepted only as 7 or 30.
Markets, spreads, earnings (v2) · 17
GET/api/v2/carry.jsonCarry pairs for the `carry` intent (G7): spot↔perp and perp↔perp, NET funding APR now/7d/30d, basis, typical spread, min sizes. Fees not included. Data only.
GET/api/v2/carry/{pid}.json
GET/api/v2/dexes.json
GET/api/v2/earnings.json
GET/api/v2/earnings/{coin}.json
GET/api/v2/earnings/{coin}/reactions.json
GET/api/v2/events.json
GET/api/v2/listings.jsonNew listings (QA-PERSONAS F9): markets whose listing date is within `days` (default 30, ≤ 3650), youngest first. `all=1` returns every market with its listing date (unknown ones last). Listing date: the collector's first sighting in metaAnd
GET/api/v2/listings/new.jsonNew listings on every dex (G8): youngest first, with first trade, live OI / volume / funding and since-listing stats (core.listings.new_listing_rows). `key` is stable per listing (tg4 new_listing alert dedup).
GET/api/v2/markets.json
GET/api/v2/markets/{coin}.json
GET/api/v2/preipo.json
GET/api/v2/scanner.jsonDefault sort gap_ext (move since external pricing stopped, the weekend "Gappers" number); thin books (liquidity_tier 'thin') are left out unless include_thin=1 (QA-R3 P1-7/P1-8).
GET/api/v2/spreads.json
GET/api/v2/spreads/custom/history.jsonA UI-built spread (CONTRACT-v3 §2.3.2 custom schema) computed on the fly from stored data; no network.
GET/api/v2/spreads/{sid}.json
GET/api/v2/spreads/{sid}/history.json
Flow, whales, liquidations (v1) · 26
GET/api/flow/v1/book/{coin}.json
GET/api/flow/v1/cohorts.json
GET/api/flow/v1/cohorts/{coin}.json
GET/api/flow/v1/cohorts/{coin}/flow.json
GET/api/flow/v1/coins.json
GET/api/flow/v1/entries/{coin}.json
GET/api/flow/v1/events.json
SSE/api/flow/v1/events/stream
GET/api/flow/v1/events/{eid}.json
GET/api/flow/v1/funding.jsonDelisted HL markets are excluded unless include_delisted=1 (QA-PERSONAS F3); sort=spread (default)|hl|coin.
GET/api/flow/v1/funding/{coin}/history.json
GET/api/flow/v1/hlp.json
GET/api/flow/v1/liquidations/cascades.json
GET/api/flow/v1/liquidations/heatmap.json
SSE/api/flow/v1/liquidations/stream
GET/api/flow/v1/liquidations/tape.json
GET/api/flow/v1/liquidations/{coin}.json
GET/api/flow/v1/liquidations/{coin}/events.json
GET/api/flow/v1/series/{coin}.json
GET/api/flow/v1/status.json
GET/api/flow/v1/tape.json
SSE/api/flow/v1/tape/stream
GET/api/flow/v1/top/{coin}.json
GET/api/flow/v1/wallet/{address}.json
One address's open positions (size, entry, liquidation price from Hyperliquid, leverage, uPnL), accounts, recent trades. liq_px = HL's liquidationPx as read, null when none; nothing is ranked by distance to liquidation.
GET/api/flow/v1/wallet/{address}/events.json
GET/api/flow/v1/whales.json
Edges · 4
GET/api/edges/catalogue.json
GET/api/edges/live.json
GET/api/edges/status.json
GET/api/edges/{edge}/{subject:path}.json
Scorecards and leaderboard · 3
GET/api/social/leaderboard.json
GET/api/social/scorecard.json
GET/api/social/scorecard/{coin}.json
Fees · 1
GET/api/fees/{address}.json
Referrals (public lookups) · 2
GET/api/referrals/code/{code}.json
GET/api/referrals/statements.csvFounder only (FUNNEL_TOKEN): every referrer's statement for an ISO week. A report: the founder pays by hand.
Share cards · 1
GET/api/cards/{cid}.json
Weekend Board and reopen desk · 44
GET/api/ant.json
GET/api/board.json
GET/api/brand.json
GET/api/contest/v1/contests
GET/api/contest/v1/entry/{eid}
GET/api/contest/v1/hall-of-fame
GET/api/contest/v1/season
GET/api/contest/v1/season/{sid}
GET/api/contest/v1/{key}
GET/api/contest/v1/{key}/leaderboard
GET/api/fees.json
GET/api/growth/preipo.json
GET/api/growth/status.jsonOperational status of the growth jobs (STRESS-W4 P2-10): builder-fills ingest freshness (a stalled ingest is flagged, and fee tiers fall back to base meanwhile), partners.yaml problems, and whether the production secrets are set. Public: no
GET/api/ipo/{ticker}.json
GET/api/ipo/{ticker}/leaderboard.json
GET/api/m/{coin}.json
GET/api/partners/group/{gid}.json
GET/api/push/v1/config.json
GET/api/region.jsonThis visitor's regional risk notice (from CF-IPCountry; config/regions.yaml), or null. A disclaimer only: nothing is restricted by it.
GET/api/reopen.json
GET/api/signals/v1/crypto/digest.json
GET/api/signals/v1/crypto/funding.json
GET/api/signals/v1/crypto/funding/snapshot.jsonD3: every coin's current funding (the same rows as the bot's /funding and C3), sorted by OI, plus C3's extremes and flips.
GET/api/signals/v1/crypto/funding/{coin}.json
GET/api/signals/v1/crypto/oi.json
GET/api/signals/v1/crypto/weekend.json
GET/api/signals/v1/earnings/afterhours.json
GET/api/signals/v1/earnings/preview.json
GET/api/signals/v1/gaps.json
GET/api/signals/v1/holidays.json
GET/api/signals/v1/macro/calendar.json
GET/api/signals/v1/macro/latestFor the board's /macro page: per release kind, the latest stored print (with its HL reaction) and the next scheduled release with its status.
GET/api/signals/v1/macro/latest.jsonFor the board's /macro page: per release kind, the latest stored print (with its HL reaction) and the next scheduled release with its status.
GET/api/signals/v1/macro/prints.json
GET/api/signals/v1/macro/prints/{eid}.json
GET/api/signals/v1/macro/reactions/{eid}.json
GET/api/signals/v1/status.json
GET/api/tape/holiday.json
GET/api/tape/post/{pid}.json
GET/api/tape/posts.json
GET/api/tape/scorecard.json
GET/api/tape/week/{week}.json
GET/api/visitor.json
GET/api/weekend/v1/last.json

Versioning

Versioned paths (/api/v2/, /api/flow/v1/) only gain fields within a version; we don’t rename or remove a field, or change its meaning, inside one. A breaking change gets a new version path, and the old one keeps working for at least 90 days after the change is announced on this page. Unversioned paths follow the same rule. Ignore fields you don’t know. API description version 1.0.0.

Example

curl -H "X-API-Key: ab_…" https://kerbtape.com/api/flow/v1/events.json?min_usd=250000
curl -N https://kerbtape.com/api/flow/v1/events/stream?min_usd=250000

Historical data API key

Hyperliquid’s weekend and overnight behaviour, as tables: date-range queries, JSON or CSV, paged with a cursor. Every dataset says where its numbers come from and how they’re measured. Market level only: no wallet addresses.

DatasetWhat it holdsMethod
weekend_premiumHourly Hyperliquid price during each weekend closure against the reference close before it.premium_bp = (px / close_px − 1) × 1e4. px = the last mid snapshot of each UTC hour (or the last candle close of the hour, px_source says which). Closures come from the weekend scorecard backfill (one per market per weekend). close_px is the exchange's official close only where close_official is true (US stocks and ETFs, from a licensed end-of-day source); everywhere else it is Hyperliquid's own price at the close and close_label says so. Pre-IPO markets and weekends before a listing are left out.
overnight_premiumHourly Hyperliquid price between a weekday close and the next cash open (≤ 20 h; longer stretches are weekends or holidays and are in weekend_premium).premium_bp = (px / close_px − 1) × 1e4; px = last mid snapshot per UTC hour; session = the market's session at that hour (overnight, extended, closed …). close_px is the exchange's official close only where close_official is true (US stocks and ETFs, from a licensed end-of-day source); everywhere else it is Hyperliquid's own price at the close and close_label says so. Pre-IPO markets and weekends before a listing are left out.
boundsPer market per weekend closure: the anchor, the starting band, the furthest reachable band, re-anchors used and whether a direction hit its hard cap. path_status says whether we know.The re-anchor counters are not on-chain, so the oracle path we sampled (about 1/min) is replayed. path_status: known = sampled throughout; partial = a gap over 15 min (a re-anchor inside it can be missed); unknown = no samples at all (our collector wasn't recording): up_used/down_used/caps are null, which means UNKNOWN, not zero. Only closures since our collector started (7 Nov 2026) can be known.
reanchorsOne row per re-anchor found by replaying the sampled oracle path. Closures whose path is unknown (bounds.path_status) can't produce rows: no row there means 'not known', not 'none'.As bounds. Several re-anchors can share one sample when the oracle jumps.
reopen_gapsPer market per weekend: the reference close, last external price, final weekend price and the reopen reference, with the close → reopen gap.reopen_px = Hyperliquid's median price 5–10 min after the reopen (not the exchange's opening print); gap_bp = (reopen/close − 1) × 1e4, recomputed against the official close when close_official. Methods per price are given; incomplete rows say so. close_px is the exchange's official close only where close_official is true (US stocks and ETFs, from a licensed end-of-day source); everywhere else it is Hyperliquid's own price at the close and close_label says so. Pre-IPO markets and weekends before a listing are left out.
funding_sessionsPer market, UTC day and session: number of hourly funding prints, mean / min / max hourly rate, the summed rate (what one unit of notional paid) and the APR of the mean.Each print is labelled with the market's session over the hour it accrued (cash, overnight, extended, weekend, holiday, closed; native crypto = always). Positive = longs pay shorts. Days are whole UTC days: a `from` inside a day starts at the next day.
liq_cascadesChains of inferred liquidations on one market, each within 60 s of the previous, with at least 2 of them AND $1M in total. Market level only: no addresses.Liquidations are inferred from public position changes and trades (not an official feed). A poll-inferred liquidation is placed anywhere inside its read window when that window is ≤ 30 min (timing = window; t_start is the earliest possible time); wider windows are left out. px_start = last trade-minute close before t_start, px_end = the close of t_end's minute.
scorecard_rowsPer market per weekend: was the weekend price closer to the reopen than the close?closer_final = |weekend_final − reopen| < |close − reopen| (ties not counted); dir_ok = the weekend move had the sign of the close → reopen move. Same rows and the same Hyperliquid close as /scorecard (close_label). Pre-IPO markets are left out.
earnings_reactionsPer earnings release: the last Hyperliquid price before it and the price 5, 15, 30, 60 and 240 min after, plus the first externally-priced print and the 1/3/5-day drift.pre_px = close of the finest candle ending at or before the release; px_Nm = the same N minutes later, each only from candles no coarser than N minutes (else null); ret_Nm_bp = px_Nm / pre_px − 1. resolution = the coarsest candle used for those prices (1m or 5m from late Oct 2026; earlier releases are 15m or 1h). gap_first/drift come from the 1h/4h reaction study (daily_resolution). Hyperliquid prices only, never options data.

GET /api/data/v1/<dataset>.json or .csv with from, to (YYYY-MM-DD, UTC; to is inclusive), optional coin, limit (up to 1000) and cursor. JSON returns next_cursor; CSV sends it in the X-Next-Cursor and Link headers. Columns and sources: /api/data/v1/datasets.json.

Limits per key: 60 requests a minute, 100,000 rows a UTC day, up to 92 days per query. API pilot keys for a few desks (3 at most) get 300 a minute, 2,000,000 rows a day and 366-day queries; ask us below. History starts where our own recording starts and has gaps where it does; rows say how each price was taken, and datasets.json gives each dataset’s coverage (markets, first and last date, rows). Closes: every close is Hyperliquid’s own price at the close (close_label says how it was taken), not the exchange’s official close.

curl -H "X-API-Key: ab_…" "https://kerbtape.com/api/data/v1/reopen_gaps.json?from=2026-09-01&to=2026-11-10"
curl -H "X-API-Key: ab_…" -D - "https://kerbtape.com/api/data/v1/weekend_premium.csv?from=2026-11-01&coin=xyz:TSLA"

KerbTape fills vs arrival, 2026-10-17 to 2026-11-15

Not enough data to publish yet: this needs at least 5 traders sharing, with uploads that match Hyperliquid’s builder-fill records. Sharing is off by default; turn it on in the terminal’s Execution report, where you see exactly what is sent.

How this is measured
  • Source: terminal users who opted in to share anonymised execution aggregates. Each uploads one summary per completed UTC day: no address, coin, order id or timestamp.
  • Only uploads whose notional matches Hyperliquid's builder-fills record for that wallet and day are used; days not yet in that record wait until they are checked.
  • Taker fills (the order crossed the book): slippage = side × (fill price / arrival mid − 1) in bp, arrival = mid when the order was sent; positive = worse than arrival. Cost = slippage + Hyperliquid fee + builder fee.
  • Maker fills (the order rested): no 'cost' and no 'savings' is computed against arrival, because where a resting order fills is mostly where it was placed. Shown instead: fees and markouts = side × (mid N minutes after the fill / fill price − 1) in bp, positive = the price moved in the trader's favour after the fill.
  • Fill rate = filled notional / notional of the orders sent (sized at arrival); orders cancelled with no fill count as unfilled.
  • Windows are fixed (7 or 30 days, ending at the last completed week, Sunday UTC) and change once a week. Figures are rounded. A cell with fewer than 5 contributors, or whose change since last week (or the gap between the 7- and 30-day windows) involves 1–4 contributors, is not shown.
  • This is a description of past fills, not a forecast; there is no comparison against other venues or tools.

JSON: /api/pro/exec/report.json?window=30 (or window=7).

API keys free in beta

Without a key, each IP gets 120 requests a minute. A key gives you 1200 requests a minute on the board’s JSON API; the historical data API above allows 60 a minute per key (every response carries X-RateLimit-Limit, -Remaining and -Reset). Send it as the X-API-Key header (or ?api_key= from a browser). Keys are issued to a wallet you sign in with: no email, no card.

Volume fee tiers

KerbTape’s builder fee drops automatically with your volume: whichever is higher of your 30-day volume routed through KerbTape (from Hyperliquid’s builder-fill records) and your own 14-day Hyperliquid volume (from Hyperliquid’s public fee data). Heavy traders start on their tier from day one. No points, no rebates to chase: the lower fee simply applies to your next order. Hyperliquid’s exchange fees are separate and unchanged.

TierRouted 30 days, or your HL 14 daysStandard marketsGrowth-mode markets
BaseUnder $1.0M3bp1.5bp
ProFrom $1.0M1.5bp0.75bp*
DeskFrom $10.0M1.5bp0.75bp*

Growth-mode markets: Half rate on HIP-3 growth-mode markets from 12 Nov 2026. The Board reviews it between 10 Dec 2026 and 24 Dec 2026; until then it stays as it is, and any change is published on the pricing page. Post-only and grid orders pay 0.5bp. The full schedule, with the referral and invitee terms: pricing (/api/fees.json).

Desks we work with directly can get the builder fee waived for up to four weeks while they try KerbTape; after that the normal tier above applies, worked out from their routed or their own Hyperliquid volume, whichever is higher. Each waiver is approved one by one and shows in the fee answer as waived_until.

Check yours: https://kerbtape.com/api/fees/<your address>.json. * Hyperliquid charges builder fees in whole tenths of a basis point, so this tier is charged rounded down (in your favour), never up.

Custom tooling

Running a desk, a fund or a trading group? Tell us your workflow: the markets, the hours, what you watch and what slows you down. We build features for real desks first, and we’ll tell you honestly if something isn’t a fit.

Contact: [email protected]

Stored on our server for the founder only, never shared or used for anything else.

The KerbTape bot never messages you first. Offers come from the founder by hand. We never ask for funds, keys or seed phrases.