Workspace / Research tools
api

API data methodology

Each returned number has a scope: supplied scenario input, provider observation or a calculation derived from observed depth. These categories are kept explicit so consumers can decide whether a result fits their own use.

Provider and quote identity

The fixed adapters call public market-data endpoints on Binance, Bybit, OKX, KuCoin and Kraken. A caller chooses from enumerated providers and base symbols; it cannot supply a URL, redirect target, private exchange endpoint or arbitrary request header. The server does not hold exchange trading credentials.

Kraken instruments use USD and the other configured adapters use USDT. USD and USDT are not silently treated as interchangeable. Provider instrument identifiers are checked where the response supplies them. Unsupported, malformed, locked or crossed books are excluded rather than converted into apparent arbitrage.

Times and cache context

received_at_ms is when this API received and normalized a provider response. as_of_ms is the provider snapshot timestamp when available, or the end of the latest completed candle for history. A book without a validated snapshot timestamp has a null as_of_ms and is explicitly labeled receipt-only. Reception time is not fabricated as exchange time.

Books are cached for five seconds and candles for thirty seconds in each worker process. Responses preserve their original timestamps and report whether the response was a cache hit, its age and its cache scope. A low cache age does not itself prove that an exchange observation is fresh. Timestamped books older than sixty seconds are rejected. Candle history is rejected when its latest completed end is older than two selected intervals or five minutes, whichever is longer.

Depth and candle coverage

Order-book output is limited to the best one hundred valid levels on each side. Duplicate aggregated prices and inconsistent data are rejected. Derived liquidity uses quote-value totals inside 10, 25, 50 and 100 basis-point bands around the mid-price. The buy walk reports the visible quantity, spent quote amount, partial-fill VWAP and whether the full requested quote amount is covered.

Candles are returned in ascending order with base-asset volume and at most two hundred completed rows. Current candles are excluded; gap_count identifies nonconsecutive intervals. The service does not fill gaps with synthetic candles. A visible book can change or cancel, and historical candles do not prove that a proposed order could have filled.

Scenario conventions

Funding uses a constant rate and notional over equivalent periods, with explicit long/short cashflow signs and supplied costs. Position sizing uses the adverse stop fill plus entry and exit fees. The arbitrage model compares equal base quantities and includes both fees and fixed costs. None executes a trade or guarantees a maximum realized loss.

The AMM model is a two-token constant-product pool with the fee retained in input reserves. Health factor divides liquidation-weighted collateral value by supplied debt; zero debt returns a null ratio and a no-debt status. Staking applies fixed-rate mathematical compounding and subtracts commission from rewards at the end. These models omit protocol-specific behavior identified in their response assumptions.

Rights and availability

Public technical documentation does not automatically grant paid redistribution rights. The deployment starts with no licensed provider enabled. The operator must review applicable terms and obtain any required permission before configuring a provider for paid access.

These adapters are bounded research integrations, not a promise that every venue listing, geographic region or provider policy remains unchanged. The calculation-only service can operate while all market adapters remain disabled.

Primary documentation

Find a tool or research page