QuantConnect Cloud

What it is

The hosted product built on LEAN: a browser IDE and a cluster you rent, with QuantConnect's datasets already mounted and its brokerage connections already wired. The algorithm you write is the same QCAlgorithm subclass the open-source engine runs, and that is the point of the split — the engine is free and knows nothing, and this is the thing that supplies what the engine does not have.

For a crypto reader the boundary is unusually clean, because the missing pieces are specific and nameable. Run LEAN on your own machine and you supply every candle, every tick and every funding rate yourself, in LEAN's zip-and-CSV layout, and a perpetuals backtest with no funding-rate file loaded charges nothing and logs nothing. Run it here and the crypto price data, the crypto-futures price data and the margin-rate datasets that feed the funding model are already there. That is what the subscription is for. It is not a better engine; it is the data and the machine.

The second thing it is for is live deployment. Live trading on the cloud runs on co-located servers racked at Equinix, from a node you rent per running algorithm, against an exchange account you hold yourself.

Pricing

Five levels, and the figures below are the month-to-month rates — the billing control on the pricing page is a checkbox that starts unchecked, which was confirmed in the rendered page rather than assumed. Read on 2026-09-19:

• Free — $0. Unlimited backtesting in the cloud, all asset classes, minute/hour/daily resolution, one backtest node, one research node, 200 projects, 500MB of workspace, 32KB files, 10,000 orders and insights per backtest, community support. Zero live trading nodes and no paper trading.
• Researcher — $84 a month. Adds tick and second resolution, the LEAN CLI for local coding, API access, paper trading, two backtest nodes and two live nodes, 2GB of workspace.
• Team — $168 per user per month, minimum two users. Ten nodes of each kind, ten collaborators, 4GB.
• Trading Firm — $480 per user per month, minimum two users. Unlimited nodes, permissions control, custom LEAN versions.
• Institution — $1,272 per user per month, minimum five users. On-premises deployment, encrypted code, IP access control, FIX and prime brokerages.

Annual billing is advertised as saving $288, $768 and $2,352 per user per year on the three team tiers respectively; no saving is printed on the Researcher card.

Every figure is a "recommended setup", not a list price, and the vendor says so in its own FAQ: plans are configured from seats, compute nodes, add-ons, data and support, and the number on the card is one common configuration. Add a larger backtesting node or a second live node and the number moves. Billing is per organisation, and an individual account is simply an organisation of one.

Separately there is QCC, a prepaid token at 1 USD = 100 QCC with a 10% bonus over $100, used for metered things — an SMS alert costs 1 QCC in the US and Canada and 10 QCC internationally. Alternative data, cloud optimisation and email support are all "Add On" line items on every paid tier rather than included.

Data & coverage

Historical crypto, from the dataset pages rather than the marketing page. Binance spot price data starts July 2017 and is delivered at any frequency from tick to daily. Binance crypto futures and Bybit crypto futures both start August 2020, likewise tick to daily, at 806 and 631 pairs.

The funding data is the part worth paying for. The Binance Crypto Future Margin Rate dataset covers the same 806 pairs from August 2020 on a daily update, built from Binance's own API, and the vendor describes it as "an important companion" to the price data because it supplies the margin interest the engine needs. Bybit has an equivalent. This is precisely the file whose absence makes a locally-run perpetuals backtest silently funding-free.

Live feeds and their sources. Crypto spot is gathered by WebSocket from Binance and Binance US, Bitfinex, Bybit, Coinbase and Kraken. Crypto futures come from Binance, Bybit and dYdX. Bars are built from ticks with microsecond timestamps, and the docs note that a tick can land in a different bar live than it did in backtest — the difference is small, and it is the kind of thing worth knowing before you spend a day reconciling two equity curves.

There is a geography constraint and it is not buried. QuantConnect's own documentation states that Binance and Bybit restrict access from the United States, and that you must use the Europe (WAW) region both to trade with them and to receive their data. For anyone building a Binance-based strategy this decides where the algorithm runs.

Integrations

Crypto brokerages on every paid tier, from the pricing matrix: Binance (crypto and futures), Bybit (crypto and futures), Kraken, Coinbase, Bitfinex, and Alpaca for US crypto. dYdX has a complete cloud brokerage page in the documentation — margin accounts, API trading keys, testnet keys for paper trading — and does not appear in the pricing matrix at all.

The non-crypto brokerage list is long and is why the platform exists in its present form: Interactive Brokers, TradeStation, Tastytrade, Charles Schwab, Tradier, Clear Street, Webull, Public, SS&C Eze, Trading Technologies, Wolverine, Bloomberg EMSX and FIX connections at the top tier.

Work happens either in the browser IDE or locally through the LEAN CLI, which is a QuantConnect- published Python package — lean 1.0.229 on 2026-08-28, with 1.0.228 in August and 1.0.227 in June. The CLI is a paid-tier feature, as is API access, so the free plan is a browser-only product. Live algorithms can send email, webhook and Telegram notifications, rate-limited per tier, with SMS billed in QCC.

Limitations

The free plan cannot trade, and that is the real shape of it. Zero live trading nodes, no paper trading, no LEAN CLI, no API, no tick or second resolution, 10,000 orders per backtest. "Unlimited backtesting" is true and is not the same as a usable free tier: it is an evaluation environment for minute-bar strategies you run in a browser.

The price on the card is not the price. Because every tier is a configurable recommended setup and node pricing lives behind a client-side "Customize Plan" flow, the cheapest configuration that actually runs your workload cannot be read off the page. Budget the tier as a floor.

The vendor's pages disagree with each other about crypto coverage. The documentation says the Binance spot dataset covers 3,651 pairs; the dataset listing page for the same dataset says 2,684 and credits it to CoinAPI rather than to QuantConnect. dYdX has a cloud brokerage page and no row in the pricing matrix. Neither discrepancy is fatal, and both mean a claim on this platform should be checked against the page the vendor has to keep current — the docs and the pricing matrix — and not against the dataset marketing page.

No crypto options, anywhere. This is inherited from the engine: LEAN models equity, index and futures options thoroughly and has no crypto option security type at all. No dataset here changes that.

No on-chain anything. dYdX is reachable, and it is reachable as an exchange API with an order book, not as a chain. There is no pool state, no swap routing, no wallet and no on-chain data in the platform. A strategy that needs to read a contract is not a strategy this runs.

One live node per running algorithm. Two strategies in production means two live nodes, which is a per-algorithm cost rather than a per-account one and is the line item most likely to surprise somebody pricing the Researcher tier against a portfolio of ideas.

You still need the exchange account. QuantConnect holds no funds and requires no identity documents of its own; every live deployment runs on API keys from an exchange that will have asked for them. The platform's kyc_required being false is a fact about the platform and not about the route to trading with it.

Alternatives

The first comparison is with the engine underneath, and it is genuinely a choice rather than an upsell: if you already have tick data and funding rates, running LEAN yourself costs nothing and gives you the same simulation. What you are buying here is not having to assemble that, plus a place to run it afterwards.

The second is against the crypto-native frameworks in Backtesting & Research Libraries. A platform whose crypto surface is one asset class among equities, options, futures, forex and CFDs will always know less about perpetuals than something written only for them — and will always know more about portfolio accounting across everything else. Pick by whether your strategy has one leg or two.