QXX.fi
All posts
EngineeringMay 28, 2026 · 7 min read

Designing a streaming API for financial data

Ordering, replay, backfill and filtering are table stakes for real-time finance. Here are the tradeoffs behind the QXX stream.

A streaming API for markets has to answer questions a generic pub/sub never does. What happens on reconnect? Are events ordered? If a consumer falls behind, do they lose data or catch up? Get these wrong and downstream systems quietly diverge from reality.

We chose at-least-once delivery with idempotent event keys, which gives consumers effectively exactly-once processing without the coordination overhead. Within a subscription, events are strictly ordered, so a consumer can maintain state deterministically.

Filters run server-side. A desk that only cares about large trades on a few chains shouldn't pay to receive — or parse — everything else. Pushing the predicate to our edge cuts bandwidth and latency at the same time.

Finally, the same normalized events that power the stream also power the historical store. That means a signal developed against history behaves identically live, because it is literally the same data arriving in a different shape.

Get started

Build on QXX.

Get API access to start querying, or book a demo to see the platform on your data.

Usage-based pricing · Free tier · SOC 2–aligned