NautilusTrader

What it is

An event-driven trading engine with a Rust core and a Python strategy API. The selling point is parity: the same Strategy subclass runs against the backtest matching engine and against a live venue, so there is no rewrite between research and production.

It is a library, not a product — no chart, no screener, no GUI. You write Python, point the engine at a Parquet catalog or an adapter, and read the result back as a pandas report. Python 3.12–3.14.

The project is mid-rewrite, and that decides which version you get. 1.231.0, released 2 August 2026, is the last feature release of the Cython v1 line and the last stable release of anything; its branch has taken no commit since. Everything now lands on the Rust/PyO3 v2 line, which is at 2.0.0rc4 of 2 September 2026 — the twenty most recent commits on develop all fall on 13 and 14 September 2026. So the cadence is fast and the stable build is frozen at the same time.

Pricing

The engine costs nothing and is not crippled. Nautech Systems, a self-funded Sydney company, monetises around it. Pro is a subscription of Docker images adding dashboards, zero-copy IPC, pre-built execution algorithms (VWAP, POV, Iceberg, Sniper) and a risk engine — waitlist-only, no published price. Cloud is announced but "under active development". Institutional is a support contract you qualify for by trading size, not by asking.

Data & coverage

None of its own. Bars, ticks and order-book deltas come from whatever you connect — Databento for US equities and CME futures, Tardis for crypto history, or the venue's own feed — and the data bill is yours.

Integrations

Nineteen adapters, all marked stable, and the balance matters. Fourteen are crypto: Binance, Bybit, OKX, Coinbase, Kraken, BitMEX, Deribit, dYdX, Hyperliquid, Lighter, Lighter on Robinhood, Derive, plus Blockchain and Tardis for data. The other five are Interactive Brokers, Databento, Betfair, Polymarket and AX, a Bermuda-licensed derivatives exchange listing perpetuals on FX, equity, metal, energy and treasury underlyings. Nine order types, emulated locally where a venue lacks them, plus Black-Scholes Greeks and option chains.

That nineteen is the documented count, which tracks v2. The 1.231.0 wheel ships fourteen adapter packages: Coinbase, Blockchain, Derive, Lighter and Lighter on Robinhood are in the v2 tree only. Nothing non-crypto is affected — Interactive Brokers, Databento, Betfair, Polymarket and AX are in both.

Limitations

• Cash equities and listed futures execution means Interactive Brokers or nothing. Databento is data-only; AX lists perpetuals on equity and commodity underlyings, not the instruments themselves; everything else is crypto, sports betting or prediction markets.
• LGPL-3.0, not MIT. Linking is fine; shipping a modified engine inside a closed product is not, without publishing the changes. One wrinkle if a lawyer is reading: 1.231.0 declares LGPL-3.0-or-later, while the v2 tree's pyproject.toml and Cargo.toml both say LGPL-3.0-only. The LICENSE file is the plain LGPLv3 text in both.
• Neither line is a comfortable live-trading choice today. The migration guide gives develop_v1 critical security backports only, for about three months from the v2 cutover, and the README says release candidates are "not recommended in production environments, such as live trading controlling real capital". Both packages import as nautilus_trader, so they cannot share a virtual environment.
• Porting to v2 is not free. Module paths moved and the strategy callbacks were renamed — on_quote_tick to on_quote, on_order_book_deltas to on_book_deltas, and so on down a table in MIGRATION_V2.md.
• No dashboard, monitoring UI or alerting in the open-source build — that is what Pro sells, and the Pro page still footnotes its own dashboard as under development.
• Steep for a first framework, and a Rust toolchain to build from source.

Alternatives

Backtrader and VectorBT if you never intend to route an order. Freqtrade or Hummingbot for crypto-only automation with far less ceremony. QuantConnect LEAN for the same backtest-to-live parity with hosted data and execution attached. The rest of the backtesting and algo frameworks are here.