Frequently asked questions.

A single source for crypto market data. We capture spot, perpetuals, funding and full-depth order books from each exchange at the source, normalize them into one schema, and deliver them as a historical archive and a live stream. L2, L3 and L4 depth are offered wherever a venue publishes the data.

Not yet. We are in pre-registration while we build the capture and normalization pipeline. We have recorded Hyperliquid internally from Tokyo and Ireland since November 2025, and that archive becomes the first product: early access for pre-registered partners is targeted for Q1 2027 and general availability for Q2 2027. Pre-register to get early access and pre-sale pricing.

Five people. Finance Data Cloud, Inc. was founded in 2025, is registered in Delaware and works remote-first, with our founder in New York and the team in London, Lisbon, Berlin and Singapore. A pre-seed round from angel investors who work in trading and market infrastructure funds the work. Meet everyone on the team page.

Hyperliquid is Phase 1, in 2027. Every other venue on our list (Binance, Bybit and OKX in Phase 2, Coinbase, Deribit, Kraken and Bitget in Phase 3, and KuCoin, Bitfinex, Bitstamp, dYdX and others in Phase 4) is planned, subject to a data agreement with the exchange. The coverage page shows what we plan to carry for each venue and the roadmap has the phase years. To ask for a venue we don't list, use the venue request form.

Chris Smith (chris@financedatacloud.com) or Daniel Garcia (daniel@financedatacloud.com). We sign data agreements directly with each exchange, follow its licensing terms and redistribution limits, and credit it as the source. More on the coverage page.

L2 is aggregated depth: total size at each price level. L3 is every order individually, as add, modify, cancel and fill events, so you can rebuild queue position. L4 is L3 plus participant attribution, where the venue publishes it, such as the on-chain user address on every Hyperliquid order. L4+ adds cross-venue attribution from our blockchain intelligence. See order books for examples.

Every message carries two: ts_exchange, the venue's own timestamp, and ts_capture, when our collector received it, from a PTP-disciplined clock. Both are int64 nanoseconds since the Unix epoch, UTC. Venue precision varies from milliseconds to nanoseconds and is listed on each venue page. Details are in timestamps and clocks.

Every venue is captured independently from two regions, and the two streams are merged by venue sequence number, so a disconnect in one region is usually covered by the other. Any gap that survives the merge is written to the gaps table with its start, end and reason. We never fill gaps silently. Past incidents are on the status page and the method is in data quality.

For Hyperliquid, every dataset starts in November 2025, when our capture began. For every other venue, the history available depends on the agreement with that venue.

One scheme across venues: spot BTC-USDT, perpetual BTC-PERP (USDT-margined linear by default), BTC-USDC-PERP, inverse BTC-USD-INV-PERP, dated future BTC-27MAR27, option BTC-27MAR27-90000-C. The exchange's native symbol stays in venue_symbol. Full rules in symbology.

Derived fields computed from full-depth data on one clock (ts_capture) and published on a 100 ms grid: multi-level order flow imbalance, VPIN, queue dynamics, cancel-to-fill ratio, book resiliency, price impact, markouts and participant flow. In our internal study on Hyperliquid BTC, ETH and SOL perpetuals, adding them raised out-of-sample directional hit rate from 51.6% to 54.9%. Read the study for the method and its limits.

Yes. Every message is stored untouched before normalization in the raw tier, and can be requested next to the normalized tables. Raw access is included on Enterprise and available on request on Desk.

Four ways, all on the same schema: the Python SDK (pip install fdc), the REST API at api.financedatacloud.com/v1, the WebSocket stream at stream.financedatacloud.com/v1, and bulk S3 with one Parquet file per symbol per UTC day. With live=True, the SDK backfills history and then switches to the live stream without a gap. See the Python SDK, REST, WebSocket and S3 docs.

The v1 spec is a draft, published in the docs so partners can build against it now. Endpoints go live with Phase 1. Test keys start with fdc_test_ and return a 24-hour sample; they are issued to pre-registered partners ahead of early access.

Parquet, CSV and newline-delimited JSON. Prices and sizes are decimal128(38,18) in Parquet and strings in JSON, so no precision is lost to floating point.

Our target is under 5 ms of added latency, median, from capture to delivery in the same region. Enterprise customers can take the stream over a Tokyo cross-connect. REST rate limits are 10 requests per second on Researcher, 50 on Desk and custom on Enterprise; see rate limits.

Researcher is $650 a month at list, Desk is $2,400 and Enterprise is custom. Every plan includes each new venue that goes live under a data agreement, at no extra cost. The comparison table lists every feature.

40% off list for the first 12 months, for pre-registered partners who sign before general availability. That makes Researcher $390 a month and Desk $1,440. Pre-registering is free and non-binding.

Yes: 80% off Researcher, which is $130 a month, for accredited university research using historical data. Apply through pre-registration with a university email address.

Researcher and Desk are licensed for internal use only, and that rules out redistribution of any kind, non-commercial included. Enterprise plans can include redistribution licenses. The terms are in the data license and the terms of service.

Keys are shown once at creation and stored only as salted hashes. Each key can be scoped to datasets and venues, restricted to an IP allowlist on Desk and Enterprise, rotated without downtime and revoked instantly. Details in authentication.

Not yet. We are preparing for a SOC 2 Type I audit, targeted for 2027, with Type II to follow. Until then we answer security questionnaires directly; write to daniel@financedatacloud.com.

Email daniel@financedatacloud.com. We acknowledge reports within two business days and do not pursue researchers who act in good faith. The full policy is on the security page.