One symbol or a hundred, a REST poll or a pushed stream, the last tick or the last thousand candles — the same key, the same symbol names and the same JSON envelope across all of it. Every response stamps when the data actually changed.
{ "success": true, "data": { "symbol": "EURUSD", "bid": 1.08432, "ask": 1.08445, "spread": 13, // points, never null "spread_pips": 1.3, "timestamp": "2026-09-18T09:41:07Z" } }
A cache TTL says how long a value may be served before it is re-read. It says nothing about how often the value changes. Those are different numbers here, and only one of them is worth building against.
A batch arrives carrying bid and ask for every configured symbol plus the OHLCV and indicator block for whichever timeframes are due. That interval is the ceiling on how new any number can be, on every surface — REST, stream or history.
Every payload carries its own age: timestamp on a quote,
updated_at on an indicator or screener response, time on a
streamed quote. Poll on a change in that value rather than on a fixed interval and you
stop paying quota for repeats.
timestamp — quote surfacesupdated_at — indicator surfaces| Timeframe | Published | Cache TTL |
|---|---|---|
| M1 | every 60 s | 180 s |
| M5 | every 60 s | 180 s |
| M15 | every 600 s | 900 s |
| M30 | every 600 s | 900 s |
| H1 | every 600 s | 900 s |
| H4 | every 1,800 s | 2,700 s |
| D1 | every 1,800 s | 2,700 s |
Spot FX is 24/5, not 24/7. The trading week opens at the Sydney session open on Sunday and closes at the New York session close on Friday; Saturday is always closed. Both boundaries are derived from IANA time zones rather than fixed UTC hours, so they move correctly across daylight-saving changes instead of being an hour wrong for roughly seven months of the year.
Weights are the real usage multipliers, scopes are the permissions your key must carry, and the plan column is the gate enforced inside the route. Pick the cheapest surface that answers your question — the difference between a batched quote and a per-symbol loop is a hundredfold.
| Surface | Returns | Weight | Key scope | Plan | Behaviour worth knowing |
|---|---|---|---|---|---|
| GET /v1/quote | One symbol, latest bid/ask/spread | 1× | quotes | Every plan | Answers the canonical symbol with bid, ask, spread, spread_pips and timestamp. |
| POST /v1/quotes | Up to 100 symbols, chosen fields | 1× | quotes | Every plan | One request of quota for the whole batch. A malformed name inside the batch is skipped like a not-found symbol rather than failing the other 99. |
| GET /v1/ohlc | Up to 1,000 completed candles | 2× | historical | Any paid plan | Candles carry time, open, high, low, close and volume only. Bid, ask and spread are not on a candle — they are on the quote endpoints. |
| WS /ws/v1/quotes | Pushed bid/ask per subscribed symbol | unmetered | API key or monitor token | Starter, Pro, Enterprise, Monitor | Not billed on any plan: credits_per_push is 0 everywhere. Access and the symbol ceiling are what the plan buys — 5 symbols on Starter, 20 on Pro, unlimited on Enterprise. |
| GET /v1/ticks | Raw bid/ask ticks, 1-hour window | 3× | historical | Pro and Enterprise only | from and to are both required and the range is capped at one hour. The row cap is a hard LIMIT 50000 in the query, not a parameter you can raise. |
The envelope is { "success": true, "data": { … } }. All
six fields are always present.
| Field | Type | What it carries |
|---|---|---|
| symbol | string | Canonical symbol, always. |
| bid | number | Bid, from our data. |
| ask | number | Ask, from the same quote. |
| spread | number | Spread in POINTS. Coerced to 0 rather than null when it is missing, so the field is always a number. |
| spread_pips | number | The same spread converted to pips using that symbol’s digit count. null when the point spread is 0 — a zero spread is not converted. |
| timestamp | string | When our data last updated this symbol. This is the freshness figure that matters, not a cache age. |
A candle is not a quote. /v1/ohlc returns
data.candles — an array of
{ time, open, high, low, close, volume } — plus
count and retention. There is no bars key, and no
bid, ask or spread on a candle.
A key in X-API-Key and a symbol is a working call. Everything else is
choosing the surface that matches the question — and, for a poller, watching the
timestamp instead of the clock.
fields trims the payload, not the price. limit on candles is 1 to 1000, validated not clamped. EURUSD. # one symbol curl -H "X-API-Key: YOUR_API_KEY" \ "https://tickatlas.com/v1/quote?symbol=EURUSD" # up to 100 symbols for ONE request of quota - pick your fields curl -X POST -H "X-API-Key: YOUR_API_KEY" \ -H "Content-Type: application/json" \ -d '{"symbols":["EURUSD","GBPUSD","XAUUSD"],"fields":["bid","ask","spread_pips"]}' \ "https://tickatlas.com/v1/quotes" # completed candles - the key is "candles", not "bars" curl -H "X-API-Key: YOUR_API_KEY" \ "https://tickatlas.com/v1/ohlc?symbol=EURUSD&timeframe=H1&limit=10"
# poll on the INGEST cadence, not on a cache TTL. # each quote's timestamp tells you when our data last updated it; # polling faster than our data updates spends quota for a repeat. import time, requests H = { "X-API-Key": KEY } URL = "https://tickatlas.com/v1/quotes" seen = None while True: d = requests.post(URL, headers=H, json={"symbols": WATCHLIST}).json()["data"] latest = max((q["timestamp"] for q in d["quotes"] if q.get("timestamp")), default=None) if latest != seen: seen = latest handle(d["quotes"]) # d["count"] rows, d["not_found"] names the rest time.sleep(30) # our data updates about every 60 s
spread is the raw point value from our data, and it depends on
that symbol’s digit count — 13 points on a five-digit pair is not 13 points on a
three-digit one. spread_pips is the same spread normalised through the
symbol’s digits, which is the figure that means the same thing across instruments.
/ws/v1/quotes — live, and free to run.Four client actions, seven server message types, and no quota accounting anywhere in the path. Streaming is unmetered on every plan that can reach it; what the plan buys is access and the number of symbols you may hold open at once.
// 1. connect, then authenticate within 10 seconds → {"action": "auth", "key": "tk_…"} ← {"type": "authenticated", "plan": "pro", "max_symbols": "20", "credits_per_push": 0, "max_connections": 2} // 2. subscribe by canonical name → {"action": "subscribe", "symbols": ["EURUSD", "XAUUSD"]} ← {"type": "subscribed", …} // 3. one message per symbol per update, about every 60 s ← {"type": "quote", "symbol": "EURUSD", "bid": 1.08432, "ask": 1.08445, "volume": 1184, "time": "2026-09-18T09:41:07Z"} // 4. keepalive every 30 s, or send {"action":"ping"} yourself ← {"type": "heartbeat", …}
| Plan | Stream | Symbols | Cost per push |
|---|---|---|---|
| Free / Trial / Demo | no | — | — |
| Pay-as-you-go | no | — | — |
| Starter | yes | 5 | 0 |
| Monitor | yes | 10 | 0 |
| Pro | yes | 20 | 0 |
| Enterprise | yes | unlimited | 0 |
Unmetered since 2026-07-30. Pushes used to be billed per symbol per update, which meant a dashboard left open quietly drained a daily quota through a channel that writes no usage log line. The per-push charge is now zero on every plan and billing happens only on explicit REST calls. Push activity is still counted, in a separate unbilled counter shown on your dashboard.
M1, M5, M15, M30, H1, H4, D1, defaulting to H1. The same allow-list guards
/v1/ohlc, /v1/indicator, /v1/multi and
/v1/summary, and anything outside it is a 400 carrying the valid set —
never a silent fallback. W1 is accepted by /v1/heatmap alone; nothing above
D1 is accepted anywhere else.
History depth is set per timeframe, never per plan. Every tier reads the same
window, and each /v1/ohlc response repeats it back in
data.retention — so the live figure is always one call away rather than
something to trust from a page. See the full schedule.
Every surface answers under the instrument's canonical name, and each one makes a deliberate choice about how its data is shaped.
Send EURUSD; the response echoes the canonical name.
Candle and tick queries keep one row per timestamp, so a series never repeats a timestamp and limit is not quietly divided.
Each incoming record is checked against that symbol’s last cached mid, with a percentage threshold per asset class and a separate guard on an absurd spread. A rejected record is skipped; the rest of the batch lands.
A background job walks stored candles, learns each symbol’s own trading hours from its history, and queues a backfill for anything missing — so a closed market is not mistaken for a gap.
Everything else on the platform is computed from our data, on the same symbol names, behind the same key. Weights are the real usage multipliers.
Quotes work on every plan. Candles (/v1/ohlc) need a paid plan, and
pay-as-you-go counts as paid for endpoint access. The two gates worth planning around are
the WebSocket, which starts at Starter, and tick history,
which is Pro and Enterprise only.
The shape that keeps recurring: take the expensive, wide payload once, then keep the moving edge fresh from the cheapest surface that carries it.
One POST /v1/quotes covers a hundred-symbol board for a single request of quota, and every row carries its own timestamp so a stale instrument is visible rather than silently frozen.
Draw the history from /v1/ohlc once, then keep only the right-hand edge moving from the WebSocket. The candle endpoint is 2× and the stream is free, so the cheap surface carries the frequent work.
spread and spread_pips answer "what would this trade cost right now" without a second call.
How fresh the numbers really are, what the stream costs, what a candle does and does not carry, and whether a quote is ever an average.
As fresh as the last update of our data, which arrives about every 60 seconds covering every configured symbol and timeframe. Every response tells you exactly when that was — timestamp on a quote, updated_at on an indicator payload — so you never have to infer freshness from a cache figure. Cache TTLs exist to survive a gap in our data, not to describe how new the numbers are: they are 180 seconds on M1 and M5, 900 on M15 through H1 and 2,700 on H4 and D1, deliberately longer than the interval at which each timeframe is published.
Both. /ws/v1/quotes is live and pushes a quote message per subscribed symbol whenever our data updates. It is available on Starter, Pro, Enterprise and the Monitor plan, and it is unmetered — credits_per_push is 0 on every plan, so a tab left open all day costs nothing against your quota. The ceiling is the simultaneous symbol count: 5 on Starter, 10 on Monitor, 20 on Pro, unlimited on Enterprise, with at most two concurrent connections per account.
No. A candle from /v1/ohlc is time, open, high, low, close and volume. Bid, ask and spread live on /v1/quote and POST /v1/quotes for the live picture, and on /v1/spread for the statistics. Joining them is a timestamp join you do client-side.
Because spot FX is a 24/5 market, not 24/7. The week opens at the Sydney session open on Sunday and closes at the New York session close on Friday, with Saturday always closed — and both boundaries are derived from IANA time zones, so they shift correctly with daylight saving instead of drifting by an hour for seven months of the year.
No. bid, ask and spread are one quote from our data, never an average, and candle and tick history keep one row per timestamp.
Coverage follows our data, so a fixed number on a marketing page would be stale the next time a symbol is added. GET /v1/symbols returns the live list with pagination up to 500 rows a page and filters for forex, commodities, indices, crypto and stocks, and GET /v1/symbols/{symbol} returns the detail for one.
Every incoming record is compared against that symbol’s own last cached mid and skipped if the move exceeds the threshold for its asset class — 5% for FX majors, 15% for gold, 25% for oil, 10% for indices, 50% for the large crypto pairs — plus a separate rejection for a quoted spread wider than 10% of mid. It is an outlier filter on ingest, and the thresholds are deliberately loose: they exist to catch broken data, not to smooth real volatility.
A background scan walks price_data every two minutes, learns each symbol’s own trading profile from its history rather than assuming a schedule, and writes any hole it finds to a backfill queue. Because the trading profile is learned, a market that is legitimately closed is not reported as a gap.
A quote is 1× whether it is one symbol or a hundred in a batch. A candle request is 2×. A tick request is 3× and is Pro or Enterprise only. WebSocket pushes are 0×. There is no free tier and no self-serve trial: every account starts on pay-as-you-go with $2.50 of prepaid credit, no card and no overage.
A hundred symbols for one request of quota, a thousand candles for one (two weighted units), and a push stream that costs nothing to keep open. Start on your $2.50 of credit, no card.