for ChatGPT, Claude, Perplexity…
DEVELOPER

Downloading Hyperliquid data

· 7 min read

"Download Hyperliquid data" means two different jobs, and they have different answers: getting your own history out (taxes, accounting, reconciliation) and getting everyone's history (research, backtesting, analytics). The API covers the first; for the second, Hyperliquid publishes raw bulk data on requester-pays S3 — the same feed its own nodes write.

Key facts
  • Your own history: api.hyperliquid.xyz/info — userFillsByTime, userFunding, userNonFundingLedgerUpdates. No key; paginate forward; details in the API behavior guide.
  • Bulk fills: s3://hl-mainnet-node-data/node_fills_by_block (API format, July 2025 →) and node_trades (older schema, Mar–Jun 2025).
  • Non-trade events — funding, staking, transfers: s3://hl-mainnet-node-data/misc_events_by_block.
  • L2 order book snapshots + asset contexts: s3://hyperliquid-archive (≈monthly uploads).
  • All buckets are requester-pays — you need AWS credentials and pay egress. Fills data is ~370 GB+ lz4-compressed.

Route 1 — the info API (your own history)

POST https://api.hyperliquid.xyz/info with an address returns that wallet's fills (realized PnL, fees, liquidation flags), funding debits/credits, and ledger movements. Free, keyless, and complete enough for tax reconstruction — but bound by the rate-limit weight budget and a forward-pagination recipe covered in measured API behavior. For a year-grouped result without scripting, the scanner runs this route in your browser.

Route 2 — the S3 buckets (bulk data)

Hyperliquid's official answer to large batch requests: public, requester-pays S3 buckets. You need the AWS CLI and credentials (the requester pays egress — cents per GB), plus lz4 to decompress.

Bucket / pathContentsFormat & coverage
s3://hl-mainnet-node-data/node_fills_by_block Every fill, streamed by block — API-compatible schema lz4 · July 2025 → ongoing
s3://hl-mainnet-node-data/node_trades All executed trades (older schema — participant order differs from API fills) lz4 · Mar 22 – Jun 21 2025
s3://hl-mainnet-node-data/misc_events_by_block Non-trade events: funding settlements, staking, transfers lz4 · ongoing
s3://hl-mainnet-node-data/explorer_blocks / replica_cmds Raw blocks and L1 transactions (msgpack) rmp.lz4 · ongoing
s3://hyperliquid-archive/market_data/[date]/[hour]/l2Book/[coin].lz4 L2 order book snapshots per coin lz4 · ≈monthly uploads, gaps possible
s3://hyperliquid-archive/asset_ctxs/[date].csv.lz4 Asset contexts (mark px, funding, OI) csv.lz4 · ≈monthly
aws s3 cp s3://hyperliquid-archive/asset_ctxs/20251001.csv.lz4 . --request-payer requester
unlz4 asset_ctxs/20251001.csv.lz4

Scale warning: the trade/fill datasets run to hundreds of GB compressed (~370 GB by community count). Sync by block range or date, not the whole bucket. Third-party providers (e.g. Reservoir) resell the same history if requester-pays is a blocker.

Route 3 — archive request and explorers

For your history older than the API's retrievable window, the remaining route is an archive/history request via official support — slow, but complete. For casual lookups, block explorers (Hypurrscan and others) render per-address activity in the browser; useful for spot-checking, not for bulk.

Which route for which job

JobRoute
Tax year totals for one walletScanner (API route, zero setup)
Custom export for one wallet (dev)info API + pagination recipe
Backtest fills across the whole marketS3 node_fills_by_block / node_trades
Funding rates for all marketsS3 misc_events_by_block, or API fundingHistory per coin
Order-book reconstructionS3 market_data snapshots + explorer_blocks
Your history older than the API windowSupport archive request

Just your own history?

Skip the AWS bill — the scanner pulls your fills, funding and ledger from the public API and totals them per tax year. Free, read-only.

Scan my wallet

Common questions

Where can I download historical Hyperliquid data?

Three routes: the info API (api.hyperliquid.xyz/info) for per-address history — fills, funding, ledger; the public S3 buckets for bulk market data — node_fills_by_block, node_trades, misc_events_by_block, and monthly L2 book snapshots in hyperliquid-archive; and an archive request to support for anything older than what the API still exposes.

Are the Hyperliquid S3 buckets free?

The buckets are public but requester-pays: you need an AWS account and you pay the egress cost yourself (cents per GB). The fills dataset alone is hundreds of GB — budget and filter by block/date before bulk-downloading.

Can I download all trades ever made on Hyperliquid?

Yes — that is exactly what s3://hl-mainnet-node-data is for. node_trades covers March–June 2025 in an older schema, node_fills_by_block covers July 2025 onward in API format. Community estimates put the trade data at ~370 GB+ compressed (lz4).

Primary sources

Check the linked sources for the rules, integration limits and API definitions discussed on this page. A general source does not settle every classification question or verify every third-party integration behavior.

  1. Hyperliquid docs — Info endpoint Official read-only query definitions, returned fields and documented history limits for fills, funding and ledger activity.

How this page was produced