C++20 · Header-Only · Python Bindings

The order book that
runs at nanosecond speed

Production-grade matching engine for algorithmic and high-frequency trading. Zero heap allocations. Zero virtual dispatch. Full Python API.

BBO Query 31 ns p50

Market Order 108 ns p50

Cancel 82 ns p50

Test Suite 56/56 passing

Get FastBook — from €299 View benchmarks

Code

Drop-in. Single include. No dependencies.

main.cpp — C++20

// One include. Everything you need. #include <fastbook/fastbook.hpp> struct MyListener : fastbook::BookListener<MyListener> { void on_trade_impl(const Trade& t) { printf("FILL %lld @ %lld\n", t.qty, t.price); } }; MyListener listener; OrderBook book(listener); Order ask{}; ask.id=1; ask.side = Side::Sell; ask.price = 10000; ask.qty = 100; book.add(ask);

strategy.py — Python 3

import fastbook class Strategy(fastbook.Listener): def on_trade(self, t): print(f"FILL {t.qty} @ {t.price}") def on_bbo_change(self, bid, bq, ask, aq): spread = ask - bid if spread < 5: self.fire_signal() book = fastbook.OrderBook(Strategy()) book.add(fastbook.Order( id=1, side=fastbook.BUY, type=fastbook.LIMIT, price=10000, qty=50, symbol="BTC-USD"))

Architecture

Built for the hot path

⚡

Zero heap allocations

Lock-free memory pool pre-allocates all order and node memory at startup. After init, the hot path never touches malloc or free.

◈

Cache-aligned Order struct

The Order struct is exactly 64 bytes — one cache line. The CPU fetches an entire order in a single memory operation.

∅

Zero virtual dispatch

CRTP callbacks resolve entirely at compile time. The compiler can inline your listener directly into the matching loop.

⊞

Complete matching engine

Limit, market, IOC, and FOK orders. Price-time priority (FIFO) per level. Cancel and reduce-qty in O(1).

⌬

Python bindings included

Write strategies in Python, execute in C++. Full event callbacks, BBO query, depth snapshot — all exposed to Python.

✓

56-test suite + benchmarks

Every code path tested. Nanosecond-precision latency benchmark included. Ship with confidence.

Performance

Measured. Not estimated.

Single-threaded, x86-64, GCC 13.3 -O3 -march=native. 100,000 samples. Includes steady_clock overhead (~15 ns).

Operation	p50	p95	p99	Notes
BBO query	31 ns	32 ns	33 ns	best_bid() + best_ask()
Cancel order	82 ns	167 ns	204 ns	O(1) hash map + intrusive list
Market order (1 fill)	108 ns	213 ns	270 ns	match + emit trade event
Add limit (100 resting)	111 ns	166 ns	204 ns	no match, rests in book
Depth snapshot (10 levels)	~180 ns	—	—	O(depth) map iteration

Run ./bench_latency on your own hardware to reproduce.

Pricing

One-time payment. Yours forever.

No subscriptions. No seats. Buy once, use forever. Includes 12 months of updates.

Individual

€299one-time

For independent developers and researchers

1 developer licence
Personal & research use
Full source code
C++ + Python bindings
56 tests + benchmark
12 months of updates
Not for commercial products

Buy Individual

Team

€799one-time

For small teams building internal systems

Up to 5 developers
Internal tools & systems
Full source code
C++ + Python bindings
56 tests + benchmark
12 months of updates
Internal use only

Buy Team

★ Commercial

€1,999one-time

For startups and trading firms building products

Unlimited developers
Commercial products & SaaS
Full source code
C++ + Python bindings
56 tests + benchmark
12 months of updates
Priority email support
Need source modifications? → Enterprise

Buy Commercial

Need source modification rights or custom integration? Contact us for Enterprise

FAQ

Common questions

What do I receive after purchase?

A ZIP archive containing the full source code: all C++ headers, Python bindings, unit tests, benchmarks, examples, documentation, and CMakeLists.txt. Delivered instantly via Lemon Squeezy.

Does it work on Windows and macOS?

The C++ library is standard C++20 and works on Linux, macOS, and Windows (MSVC 2022+, GCC 10+, Clang 12+). The Python extension requires a C++20-capable compiler.

Can I use this in a live trading system?

Yes, under the Team or Commercial licence. The library is designed for exactly this use case. The disclaimer in the licence applies: you are responsible for validating behaviour in your own system before deploying capital.

Is there a refund policy?

Yes — 14-day no-questions refund via Lemon Squeezy if the library does not work as documented on your platform.

Does it support FIX protocol or WebSocket feeds?

Not natively — FastBook is the matching engine layer. You parse your feed (FIX, WebSocket, REST) in your own code and call book.add() with the resulting Order struct. This keeps the library focused and dependency-free.

Will there be updates?