Market data, on record.
Mirexa keeps candles and ticks for trading and research: live as they arrive, complete in history, and quick to read back.
Why Mirexa
History and live in one feed
Subscribe to a symbol and get its past bars, then every new one as it lands, with nothing missed at the handover.
Fast where trading systems look
The latest bars, or a day of one symbol, come back in under a millisecond.
Corrections without losing history
Restate a bar or load years of history behind live data. Repaired ranges stay on record.
Knows when markets are closed
Each stream has a trading calendar, so a weekend reads as closed and an outage as a hole you can find and backfill.
Stays up when a machine doesn’t
Optional clustering keeps copies of every partition, fails over on its own and repairs damaged data from a healthy copy.
A dashboard included
Browse streams, chart them, and find and fill missing ranges from the browser.
Benchmark
Compared with QuestDB, ClickHouse, TimescaleDB, TimeBase and InfluxDB 1, 2 and 3 on one laptop, loading the same 10 million candles. Mirexa’s numbers, averaged over three runs:
Measured 2026-10-17 on a laptop that was also running other work. The other databases ran with default settings. The benchmark page has every system, every run and the caveats.
Status
In development, by one developer.
- Single node
- Writes, crash recovery, queries, corrections, subscriptions, token authentication and a web dashboard work.
- Clustering
- Replication, partition moves, failover and repair of damaged chunks are built and tested on a development branch. Off by default.
- Not yet
- Server-side rollups, clients for languages other than Go, and replication of flushed tick data. Authentication is off by default. See the known limits.