Hyperliquid Historical Data: Archive Coverage, Formats and Access

Dwellir preserves Hyperliquid node data in its source record formats and offers processed datasets for backtesting, research, and analysis. This page is the single inventory for S3 Bucket contents, formats, access, pricing, and product selection.

For sample design, manifests, timestamp semantics, TWAP joins, liquidations, account-state replay, and commercial use, see archive research and validation.

S3 bucket coverageCopy Anchor Link

Compressed size and file countCopy Anchor Link

The coverage table shows the compressed download size and file count for each available prefix. You can use these totals before buying archive access.

Sizes use decimal units: 1 GB is 1 billion bytes, and 1 TB is 1 trillion bytes. Decompressed files need additional storage.

Monthly averages use the last three calendar months with files on every date. Shorter histories use fewer months.

For a selected range, estimate volume as days / 30.5 × monthly average. Trading activity changes file sizes, so the result is approximate.

File count is the number of compressed files, not the number of blocks, events, or orders. File dates do not establish that every source record is present.

Archive formatsCopy Anchor Link

Native JSON and JSONL payloads are LZ4-compressed. Decompress .lz4 objects before parsing the record shape for their prefix. Generated order-book snapshots use <height>.json.lz4, while periodic ABCI state uses <height>.rmp.lz4. The latter contains RMP-encoded node state, not JSON.

Most current *_by_block feeds use an envelope with local_time, block_time, sequential block_number, and a feed-specific events array. The replica_cmds records contain abci_block and execution responses. Its abci_block.round is a consensus round, can skip, and must not be treated as the sequential block number.

The fills history spans 3 stored prefix generations:

Observed objects for these generations overlap, so their bounds are not mutually exclusive schema cutover dates. Select the parser from the object prefix and record shape. For processed analysis, tick data converts node_fills_by_block records into CSV or Parquet. For query access, use the Hyperliquid Index fills API.

Current field-level references are available for blocks, fills, order statuses, raw book diffs, misc events, and TWAP statuses.

Each data catalog page also has an "Observed fields" tree. It lists every field seen in one real archive object for that prefix, including the periodic_abci_states key structure.

Order-book reconstruction requires a known state seed, such as an orderbook_snapshots object generated from periodic ABCI state, followed by ordered node_raw_book_diffs_by_block records. Raw diffs alone cannot establish the starting book.

Choosing a historical productCopy Anchor Link

To discover symbols before requesting data, use the Info endpoint. Perpetual names come from meta; spot names come from spotMeta.

Real-Time DataCopy Anchor Link

For live market data, see:

• Order Book Server - Real-time L2/L4 order books and trades
• gRPC Streaming - StreamBlocks, StreamFills, and more

PricingCopy Anchor Link

API plan prices, included responses, and rate limits are on the pricing page. Current export credit prices are shown on the OHLCV Data tab.

AccessCopy Anchor Link

For OHLCV data: sign up at the dashboard and visit the OHLCV Data tab to buy export credits and generate CSV or Parquet files for any market at 1s, 1m, or 5m intervals.

For S3 Bucket archival data, tick data, and other historical products:

Email: support@dwellir.com

Include in your request:

• Prefix or data product needed
• Date range
• Preferred format
• Use case description

Use the coverage table and date-range estimator to plan downloads before purchase.

One-off windows and filtered exportsCopy Anchor Link

S3 archive delivery is scoped by prefix and date. Each file covers every market, so you filter markets and exact time bounds after download. A short window still delivers the full files that contain it.

An export filtered to specific markets, decoded fields, or derived values is custom work. Dwellir quotes it separately by email. Prices and delivery times for custom work are not published.

For a custom request, also include:

• Markets
• UTC start and end, with an inclusive start and exclusive end
• Each field you need, with the source timestamp and block position you expect
• Whether you need a completeness or gap report

Check first whether a prefix stores the field. Mark and oracle prices, for example, are not stored as their own records.

After payment, Dwellir issues a read-only S3 key limited to the purchased prefixes and dates.

Frequently asked questionsCopy Anchor Link

How large are the archive files, and how many files will I receive?Copy Anchor Link

The coverage table lists compressed size and file count per prefix. Use the date-range estimator for a shorter selection. Download volume varies with trading activity.

Which Hyperliquid historical data should I use for backtesting?Copy Anchor Link

It depends on what your strategy trades on. Bar-based strategies can use OHLCV candles at 1s, 1m, or 5m. Execution backtests need fill-level tick data or the node_fills_by_block prefix. Order-level replay needs orderbook_snapshots plus node_raw_book_diffs_by_block. Their overlapping archive coverage starts on 2026-04-14.

Can I review my own closed trades against historical prices?Copy Anchor Link

Yes. userFillsByTime returns a wallet's fills back to 2025-07-27, and candle ranges return 1s and 1m candles for the same markets. The trade post-mortem guide rebuilds each closed trade from those fills, measures its best and worst move, and replays stop, target and no-adding rules on it.

How far back does Hyperliquid historical data go?Copy Anchor Link

replica_cmds starts on 2025-01-25. Fills start on 2025-03-22 in the node_trades prefix, and node_fills_by_block plus OHLCV candles start on 2025-07-27. The full per-prefix table is in S3 bucket coverage.

Does the archive include HIP-3 markets?Copy Anchor Link

HIP-3 coins appear under their dex-prefixed names, such as xyz:XYZ100. The book diffs sample includes an xyz: coin, and deployer setOracle actions are in replica_cmds. HIP-3 history starts at each market's first trade, for example 2025-10-13 for xyz:XYZ100. Ask the Dwellir team to confirm coverage for a specific dex. HIP-3 explained shows how to read HIP-3 markets live and where their candles and order book history come from.

Can Dwellir export every mark-price update for a time window?Copy Anchor Link

No prefix stores mark prices as their own records. Applied values come from replaying replica_cmds from a periodic ABCI state checkpoint. See historical mark prices for the inputs and their limits, and one-off windows to request a quote.

How much does Hyperliquid historical data cost?Copy Anchor Link

API plans start at $49/month, and 1 response = 1 API credit. OHLCV exports cost 1 export credit per market, interval, and format. S3 Bucket archive data and tick data are quoted per dataset. See Pricing.