DOCS

Options data

Full OPRA options data over REST: 12 request types from live chains to trades joined with the NBBO of that millisecond, 12 years deep.

Every US listed option, from OPRA: every trade, every NBBO quote, greeks, open interest and the joins between them. 12 request types, all under /v1/options/<type>. All of them return the same envelope except chain, which returns { symbol, count, options[] } with camelCase rows (ivError, bidIv, …).

What each package unlocks

The tiers are not three sizes of the same thing. Each unlocks the first N request types in the canonical order below, and the order is the argument: the surface, then the tape, then the joins.

The surface — what the chain looks like

The tape — what actually printed

The joins — what a print meant

Making a call

requiresOptions Core or above

curl "https://api.tick-stream.xyz/v1/options/trade_quote?symbol=SPY&exp=2026-08-21&strike=600&right=C&date=2026-07-31" \
  -H "Authorization: Bearer sk_live_…"

import requests

r = requests.get(
    "https://api.tick-stream.xyz/v1/options/trade_quote",
    params={"symbol": "SPY", "exp": "2026-08-21",
            "strike": 600, "right": "C", "date": "2026-07-31"},
    headers={"Authorization": "Bearer sk_live_…"},
).json()

# rows are OBJECTS with named fields — read them by name
for contract in r["response"]:
    for row in contract["data"]:
        aggressive = row["price"] >= row["ask"]

const u = new URL("https://api.tick-stream.xyz/v1/options/trade_quote");
u.searchParams.set("symbol", "SPY");
u.searchParams.set("exp", "2026-08-21");
u.searchParams.set("strike", "600");
u.searchParams.set("right", "C");
u.searchParams.set("date", "2026-07-31");

const r = await fetch(u, { headers: { Authorization: "Bearer sk_live_…" } });

Parameters

Which types need a date — all of them that read history

Dates are not optional on the history types, and this is the one omission worth calling out: they are not defaulted to today. Every date is YYYYMMDD (2026-07-31 with dashes is accepted too), and date is shorthand for start_date=end_date.

NDX and NDXP greeks (trade_greeks, greeks_history) are computed by us, because no index level is published with them. The underlying is the forward implied by put-call parity at the at-the-money strikes, interpolated to the second; implied vol and the greeks follow from the trade price or quote mid. Each row says so with underlying_source: "implied_forward"; a row we could not price says "unavailable" with iv_error: 1, and iv_error: 1 can also appear on an implied_forward row when the fit behind it was poor — treat both as unpriced. Checked against the vendor's own SPXW greeks, the median difference in implied vol is 0.6 vol points. Up to 31 sessions per request; greeks_history accepts expiration=* here with max_dte (0–30), and start_time/end_time are ignored — you get the whole session.

A request missing a date it needs now returns 400 bad_request with the missing parameter named, and a pointer back to this page. It used to reach the vendor without one and come back as a 502, which read like an outage on our side and was not.

The response envelope

One entry in response per contract, each carrying a contract block and a data array. Rows are objects with named fields — there is no header.format and no positional row. We forward the source body untouched rather than re-mapping forty vendor columns into names of our own, so read every field by its key and ignore key order.

{
  "response": [
    {
      "contract": { "symbol": "SPY", "expiration": "2026-08-21",
                    "strike": 600.0, "right": "CALL" },
      "data": [
        {
          "trade_timestamp": "2026-07-31T13:20:27.697", "sequence": -1534418721,
          "size": 1, "condition": 138, "exchange": 6, "price": 147.06,
          "quote_timestamp": "2026-07-31T13:20:27.292",
          "bid_size": 39, "bid": 145.68, "bid_exchange": 5, "bid_condition": 50,
          "ask_size": 18, "ask": 147.21, "ask_exchange": 9, "ask_condition": 50,
          "ext_condition1": 255, "ext_condition2": 255, "ext_condition3": 255, "ext_condition4": 255
        }
      ]
    }
  ]
}

Conventions worth knowing before your first parse

trade_quote — the one worth the upgrade

requiresOptions Pro or above

Every trade paired with the NBBO that stood in that exact millisecond. Without the pairing a print is a number; with it, it is an aggressive buy or a passive fill:

price >= ask  →  lifted the offer
price <= bid  →  hit the bid
size  >  ask_size  →  took more than was displayed

That last line is why the join matters more than the trade feed alone. A 250-lot into a 120-lot offer is a different event from a 250-lot into a 5,000-lot offer, and the trade record on its own cannot tell you which one happened.

Whole-root snapshots

requiresOptions Pro or above

/v1/options/root returns every expiration of a root in one call, as a live snapshot, rather than one request per contract. It is the difference between a chain sweep that takes a second and one that takes four hundred requests.

Bulk history — the whole chain in one request

requiresOptions Core or above

The history types eod, oi, quote, trade, trade_quote and trade_greeks accept expiration=*: one request answers every contract of the root for the date window, instead of one request per expiration. greeks_history (except on NDX/NDXP, which we compute ourselves — see above) and ohlc do not — ask them per expiration with strike=* (every strike of that expiration in one request), and narrow the day with start_time/end_time when you only need an instant. This is the shape to use for backfills — a month of QQQ end-of-day (60,260 contracts) is a single call that returns in about twenty seconds:

/v1/options/eod?symbol=QQQ&expiration=*&start_date=2019-03-01&end_date=2019-03-31
/v1/options/oi?symbol=QQQ&expiration=*&start_date=2019-03-01&end_date=2019-03-31
/v1/options/trade_quote?symbol=QQQ&expiration=*&max_dte=7&date=2024-03-15

max_dte and strike_range clip the wildcard: the third example is the full front-week tape of a session — every trade with its NBBO — without pulling the LEAPS ladder along with it. Pulling seven years this way is a few hundred month-sized requests, not hundreds of thousands of per-contract ones.

Contract lists: expirations and strikes

requiresOptions Core or above

Two metadata requests, open to every options package and not counted among the request types: /v1/options/expirations?symbol=PLTR lists every expiration the root has had (listed and expired), and /v1/options/strikes?symbol=PLTR&expiration=2026-10-16 the strikes of one expiration.

A chain nobody streams yet

chain serves the live snapshot the gateway polls. A root that nobody streams is added to the poll by the first request, which answers 503 chain_loading with Retry-After: 60; the chain is there on the next poll cycle. A root that still has none after five minutes answers 404 no_chain (no listed options). For a one-off board without waiting, root (Options Pro) reads the whole root in a single request.

Every chain row carries the first-order greeks (delta, gamma, theta, vega, rho) and the second-order ones vanna, charm, vomma and veta, alongside iv and ivError — the same row on the options WebSocket channel.

Live chains over WebSocket

Cadence, stated plainly: the live chain — REST and WebSocket alike — is a full-chain snapshot (every strike, with greeks) refreshed about every 30 seconds per underlying. It is not per-second. If you need finer, you don't poll the chain faster — you use the tape: the Flow/Pro history types carry every OPRA trade and every NBBO quote with exchange timestamps, and Pro adds the intraday greeks_history series.

requiresOptions Core or above

The options channel still streams live chains with greeks by underlying — unchanged by the provider switch. Subscribe the same way as any other channel:

{ "op": "subscribe", "channel": "options", "symbols": ["SPY", "QQQ"] }

The live tape: prints with the NBBO at them

The option_trades channel is every OPRA print as it happens, each one carrying the national best bid and offer that stood at that moment. That pairing is what makes net flow readable: a print at the ask was somebody paying up, a print at the bid was somebody hitting it. A chain snapshot can only tell you volume went up, never which side lifted.

Subscribe by underlying root — the filter is applied here, so you receive only the roots you ask for. It is included from Options Flow upwards; Core streams the live chain but not the tape.

The whole tape: on Options Pro, "symbols": ["*"] subscribes to every root at once, the full OPRA band. It counts as one symbol against your allowance, and you can combine it with named roots. On Flow the wildcard is refused with a plan_required error frame; the named roots in the same request still subscribe.

{ "op": "subscribe", "channel": "option_trades", "symbols": ["SPXW", "QQQ"] }
{ "op": "subscribe", "channel": "option_trades", "symbols": ["*"] }   // Options Pro: every root

{
  "type": "option_trade",
  "symbol": "SPXW", "expiry": "2026-09-18", "strike": 6600, "right": "C",
  "occ": "SPXW  260918C06600000",
  "ts": 1789667131278, "price": 10.9, "size": 5,
  "bid": 10.5, "bidSize": 12, "ask": 10.9, "askSize": 7, "quoteTs": 1789667131000,
  "at": "ask",
  "exchange": 65, "condition": 18, "seq": -563040482
}

at is where the print landed: "ask", "bid", "mid", or an empty string when no quote for that contract had arrived yet. We do not translate it into "buy" or "sell" — at the mid there is no honest answer, and the classification is yours to make. ts and quoteTs are exchange milliseconds. Sizes and prices are the contract's, not per share.

How far back — and how fine-grained

Two axes decide whether a plan fits a backtest: depth (how many years) and granularity (end-of-day rows vs tick-level intraday). And one rule that keeps coming up in support mail, so it gets its own sentence: nothing here is forward-only. Every history endpoint serves the archive from the first day of your subscription — a key created this morning queries all of its plan's years immediately.

The contract universe itself reaches back to 2012-06-01; a request older than your package's window is clamped to it rather than rejected. Verified at the far end on 2026-08-03: SPY 132C expiring 2012-06-16, session 2012-06-11 — 2,592 prints, 2,137 distinct millisecond stamps, full OPRA fields with the NBBO beside them.