A small team, building the hard part in the open
Open Exchange exists because the core of a trading venue, the matching engine and the accounting around it, is the part everyone underestimates and nobody wants to buy as a black box.
We started in early 2026 with a simple test: could a small team build a matching engine that holds up under the load and failure stories that sink real exchanges, and prove it in public? Every release since has been that test, repeated with higher stakes: leader failovers under load, power-loss drills, ledgers checked to the last unit.
Today the stack is four open repositories under Apache 2.0: the matching cluster, the order management service, the operations gateway, and a trading interface. It reached beta in July 2026, with a frozen public API and a failover test that runs on every commit.
We fund the work by helping teams deploy and extend it. The open core stays free; that isn't a promotional line, it's the architecture of the business.
Every performance claim on this site comes from a measured run, published with the rig, the method, and the caveats. When we got a number wrong (we once over-measured by 30x), we published the correction too.
An exchange is judged on its worst day. We kill our own cluster leader mid-load on every commit, and the release notes talk about the bugs we found as much as the features we shipped.
We build the matching, accounting, and operations layer, and we say plainly what we don’t: custody, KYC, licenses. Pretending otherwise is how exchange projects fail.
The best introduction to how we think is the work itself: the engineering notes on the blog, the release history, and the code.