Use a free key to confirm the integration path, not to design an unlimited production job. Start with one current order-book request, confirm the response envelope, and preserve meta.request_id. Then choose whether the real task needs history, streaming, SDKs, CLI, MCP Server, Skill, or OpenAPI code generation.
After the starter request works, repeat the same checklist for the actual family you need: Spot pair, HIP-3 symbol, HIP-4 outcome side, or Lighter market.
Evaluate a paid plan when the workload needs more credits, request rate, concurrency, standard replay speed, or export capacity than the current account provides. Plans govern capacity and request windows. Availability is specific to the route, family, symbol, and schema; inspect the selected operation and coverage before relying on it.
Do not use a starter key as the design point for long-running production systems, broad historical backfills, high-concurrency crawlers, replay-heavy workflows, L3/L4 reconstruction, or customer-facing exports. Those jobs should start with route coverage, rate and credit limits, data-quality gates, and commercial fit.
The documented market-data scope is Hyperliquid and Lighter: Hyperliquid core, Hyperliquid Spot, HIP-3, HIP-4, and Lighter. Keep generic RPC infrastructure, onchain wallet indexing, broad CEX catalogs, and execution APIs as separate systems.
