Laevitas API

What it is

A REST and WebSocket API for crypto derivatives, plus an MCP server over the same data. It is the programmatic half of a derivatives analytics business whose other half is a web app; this card is about the API, which is bought by somebody writing code and priced nothing like the seats.

What distinguishes it from the funding-and-open-interest crowd is the options side. The v2 API exposes per-instrument implied volatility and Greeks, options trade flow, dealer gamma exposure (latest, history, strikes, term structure, regime changes), and a proprietary calibrated volatility surface you can query as snapshots, slices, term structure or strike diagnostics — with inverse and linear margin books kept deliberately separate, which is the kind of detail that tells you the surface is calibrated rather than scraped. There are also asynchronous options backtests against stored strategy templates, and a Hyperliquid HyperCore surface derived from node data: fills, liquidations, funding payments, TWAP events, resting orders and wallet positions by address.

Version 2 lives on a different host from the original API and the two are still both running, so check which one a tutorial or a wrapper is written against before you copy it.

Pricing

Two ways to pay, and neither is a developer plan in the shape most data vendors sell.

An API key is issued against the Enterprise plan at $500 a month per seat. The plan is sold as a dashboard seat with "API historical data" as one of its bullets; the older documentation says the same thing more bluntly, that an API key requires an enterprise subscription. Above it is a contact-sales Custom Enterprise tier advertising high-throughput access. No requests-per-minute figure is attached to either, anywhere.

Or x402: call an endpoint with no key, get an HTTP 402 back with a payment challenge, sign it, retry. The challenge asks for 0.10 USDC per request, payable on Base or on Solana, with a five-minute window, and it is the same 0.10 on every endpoint sampled — options, spot, perpetual order books, the volatility surface, the instrument reference list. That is a real price rather than a marketing line: it is in the response header on an unauthenticated request. WebSocket streaming is not per-message under this scheme — you buy a pass, get back a credit token, and connect with it. An hour is 0.50 USDC, 24 hours is 10 USDC, and both are quoted the same way, in the payment challenge rather than on the pricing page. Buying a pass while one is live extends it rather than replacing it.

The free and mid tiers on the pricing page are seats in the web app. They are a different purchase with a different buyer and this card does not describe them.

Data & coverage

Options venues are four, not fifteen. The marketing site's statistics strip advertises 15+ exchanges and its REST API block names five — Binance, Deribit, OKX, Bybit and Hyperliquid — but the public instrument catalogue returns maturities and strikes for exactly Deribit, OKX, Binance and Bybit, with no Hyperliquid option chains in it at all. An AEVO key is still present in that catalogue with nothing behind it. Parameter examples in the v2 schema also mention Bullish and Derive; examples in a schema are not a coverage inventory, and the catalogue endpoint says so itself. Options coverage is always thinner than spot coverage at every vendor — this one is honest about it if you read the right page.

Eight options venues were retired, each with a published last observation: Zeta stops in September 2022, Lyra and Lyra on Arbitrum in December 2023, AEVO in June 2024, PowerTrade in March 2024, BIT and Delta Exchange in April 2025, and CME options in December 2025. If you came here for CME crypto options history, the series ends and does not resume.

History. The vendor advertises 5+ years and does not publish a per-endpoint start date. What it publishes instead is better: a machine-readable coverage endpoint that answers, per selector, the earliest and latest stored timestamp, plus a completeness watermark — rows before it are provisional and both the set of rows and their values may still change. A corrective change to an already certified interval is published separately in a corrections feed. Raw ticks are the exception and the limit is hard: raw trades and raw order books for futures, perpetuals, spot and the Polymarket surface are kept on a rolling 30-day window, and a request ending before it comes back classified as outside retention rather than empty.

Granularity. Time series bucket at 1m, 5m, 15m, 1h, 4h and 1d, up to 1000 rows a page with cursor pagination. The 1m grain is a persisted aggregate of finer source events and everything coarser rolls up from it, so 1m is the floor for history — the finer events are what the WebSocket carries live.

Streaming limits are published in full, which is rarer than it should be: 200 subscriptions per connection, 20 inbound messages per second, a 24-hour maximum connection lifetime, a 2 MB slow-consumer queue before you are dropped, 5 concurrent connections per API key and 20 per IP. Every close code is documented, including the one for an expired x402 streaming pass. Budget for the daily reconnect; it is by design, not a fault.

Nothing on the marketing site or in the OpenAPI document ties resolution, depth or freshness to a tier, so there is no plan-by-plan comparison to make here — only the seat-versus-x402 split, and the one difference that is documented between them is that streaming under x402 needs a pass.

Integrations

REST over X-API-Key, native WebSocket at the v2 host, and a legacy Socket.IO v4 endpoint kept alongside it. Channels are addressed as market, exchange, instrument and timeframe, where market is one of perpetuals, futures, options or spot.

The MCP server is a first-party one on the API host, streamable HTTP, with OAuth 2.1 and PKCE required; published client identity is explicitly unsupported, so a new connector has to register itself automatically, though manual registrations that already work are left alone. Setup guides cover Claude.ai connectors, Claude Desktop, Claude Code, Cursor, VS Code, Windsurf, Codex CLI, n8n and ChatGPT actions via the OpenAPI schema. Note the tool count disagrees with itself: the MCP page says 80+ tools, the marketing page still says 20+.

No first-party SDK is named in the OpenAPI document or on any of the developer pages. Plan on an HTTP client.

Limitations

• No published REST rate limit. Not on the marketing site — whose plans and API sections are two anchors on a single page — and not in the OpenAPI document. The gateway returns X-RateLimit-* headers, so you discover your ceiling at runtime. For a $500-a-month product that is a gap, and it makes capacity planning a support conversation.
• The entry price for a key is $500 a month. There is no $29 developer tier. x402 fills that gap only if paying per call from a funded wallet is acceptable to you — at 0.10 USDC a request, a backfill that pages through a year of hourly candles is not cheap, and an agent left looping is a bill.
• The MCP server needs an Enterprise key, so the cheapest way in — x402, no key — does not open it. Pay-per-request is advertised as working with MCP tool calls; the setup page asks for a key regardless.
• Raw trades and raw order books vanish after 30 days. If your research needs tick data older than a month, this is not the archive for it, whatever the 5+ years on the front page suggests.
• 1m is the finest historical bar, and it is an aggregate of finer events rather than the events themselves. Sub-minute history is not on offer through REST.
• Options history is venue-limited and shrinking at the edges — eight venues retired, CME options most recently. Coverage that ends is coverage you have to notice; to the vendor's credit the changelog endpoint publishes the terminations with timestamps instead of letting the series just stop.
• The vendor's own pages disagree on the MCP tool count, and the site footer still reads 2025. Small things, but they are the things that tell you which page is maintained.

Other options for this job sit under crypto market data APIs.

Alternatives

Laevitas Platform is the same vendor's other product and the first thing to rule in or out: if the surface and the screens are for you to look at rather than for your code to read, that card is 50 dollars a month and this one is 500.

Amberdata is the closest institutional substitute with options data attached, and it will sell one exchange on a card. CoinGlass is cheaper and covers funding, open interest and liquidations across thirty venues — with no implied volatility and no Greeks anywhere, which is precisely the gap this card fills. Coinalyze is the free version of that same funding-and-open-interest job.

If you would rather compute the surface yourself, Tardis.dev sells the raw options chains to compute it from, and the work is then yours.