Overview
NautilusTrader is an open-source, production-oriented algorithmic trading platform focused on performance, correctness and parity between research and production. It combines a Rust/Cython core with Python-native APIs so quantitative researchers can write strategies in Python and run the same code unchanged in live trading.
Key characteristics
- Core implementation: Rust (async networking with tokio) and performance-sensitive bindings via Cython / PyO3. Python-first API surface.
- Execution model: Event-driven engine supporting tick/bar/order-book data with nanosecond resolution for realistic backtests.
- Parity: Identical strategy implementations for backtesting and live deployment to reduce reimplementation risk.
- Precision modes: Support for standard (64-bit) and high-precision (128-bit) numeric modes to handle diverse asset classes and pricing scales.
- Order & execution features: Advanced order types and instructions (IOC, FOK, GTC, GTD, AT_THE_OPEN/CLOSE), execution flags (post-only, reduce-only, iceberg), and contingency orders (OCO, OUO, OTO).
Integrations & extensibility
NautilusTrader is adapter-driven: any REST or WebSocket venue/data provider can be integrated via modular adapters. The project ships guides and adapters for multiple exchanges and providers (Binance, Bybit, Coinbase, Interactive Brokers, Databento, Tardis, Betfair, etc.). The message-bus and cache abstractions let users extend or replace components for custom workflows.
AI & research use
Although not an ML model itself, NautilusTrader is explicitly "AI-first":
- The backtest engine is designed to be fast and deterministic enough to run large-scale experiments and train AI trading agents (e.g., reinforcement learning, evolutionary strategies).
- Its parity between backtest and live reduces sim-to-prod friction for models trained in research environments.
Deployment & developer workflow
- Packaging: Official wheels are published on PyPI and a Nautech package index; nightly/develop wheels are available for testing.
- Platforms: Linux (x86_64/ARM64), macOS (ARM64), Windows (x86_64) supported; Docker images provided (including JupyterLab examples).
- Build: Option to install from source (requires Rust toolchain, clang, uv package manager) for development or to enable platform-specific features.
Safety, testing & release cadence
- Emphasis on software correctness, type- and thread-safety from Rust; optional Redis-backed persistence for state.
- Active CI with branch-based releases (master/nightly/develop) and a roughly bi-weekly release cadence for stable releases.
- Project provides performance testing, attested build artifacts and guidance for verifying build provenance.
Community & license
- Maintained by Nautech Systems (Nautech Systems Pty Ltd). Community channels include GitHub, Discord and the official website. The project is published under the GNU Lesser General Public License v3.0.
When to use NautilusTrader
- You need a production-capable, event-driven trading engine where strategy code written in Python can run unchanged in live deployments.
- You want a backtester performant enough to support large-scale AI training loops (RL/ES) and multi-venue simulation.
- You require advanced order types, high numeric precision options, and a modular integration surface for custom adapters.
