Raw financial values preserved for research.

Captured, attributed historical data prepared manually in the formats your tooling already reads.

Book a DemoRead the docs
Your problem

The parts you keep working around.

01

Float-rounded datasets produce backtests that don't reconcile to the chain.

02

Anonymized data forecloses whole research directions: cohorts, behavior, flow.

03

Vendor exports arrive in bespoke formats that need a parser before they need a model.

The mapping

How the platform answers, point by point.

The proof

Aligned raw fields, your requested format.

Raw captured rows preserve the live field model and financial decimal strings. Manual exports can be requested in formats your tooling reads.

Archive row · trades dataset

manual Parquet / CSV / JSONL export

block_timetimestamp2026-06-09T08:14:02.418Z
coinstringBTC
pxstring67012.0
szstring0.14523
notionalstring9732.15
sidestringB
userstring0x7c4e…a3f9
closed_pnlstring182.337041
feestring2.433037
crossedbooltrue
tiduint64442191083327011

Live message · StreamSwaps

stream.hyperliquidrpc.com:443 · gRPC

"block_time": "2026-06-09T08:14:02.418Z",
"coin": "BTC",
"px": "67012.0",
"sz": "0.14523",
"notional": "9732.15",
"side": "B",
"user": "0x7c4e…a3f9",
"closed_pnl": "182.337041",
"fee": "2.433037",
"crossed": true,
"tid": 442191083327011

Hover a field. It is the same field, byte for byte, on both sides. Raw export fields stay aligned with the stream field model.

Export formatsParquetCSVJSONLManual exports are prepared after coverage and scope are confirmed.
Integration sketch

Scope first, then export.

Send the dataset, markets, range, fields, and preferred Parquet, CSV, or JSONL format. We confirm retained coverage before quoting delivery.

Read the docs 
{  "dataset": "trades",  "markets": ["BTC", "ETH"],  "from": "2026-06-01",  "to": "2026-06-08",  "format": "parquet",  "destination": "s3://your-bucket/hyperliquid/"}
Where to start

The plan that fits.

Ultra

$199/moRecommended

Recent replay and maximum throughput.

Ultra adds recent replay and higher live limits. Historical exports and custom datasets are manually scoped and quoted for the requested range.

Compare both plans

Pull a sample before you commit.

Historical exports and custom data shapes are quoted separately. Contact us with the dataset, markets, range, and format.

Book a DemoGet an API key