FaunaArchived

Fauna was sold as an API rather than as servers: you created databases in a web console and sent queries over HTTPS, and Fauna, Inc. ran the replicas in several regions. Data was JSON-like documents in collections, with schema added gradually, references between documents in different collections, and joins and projections in FQL, a query language that in its last version looked like TypeScript and could be statically typed. Schema, indexes, roles and user-defined functions were written as code in Fauna Schema Language (FSL). Databases could contain child databases, which teams used to keep tenants or staging and production apart. Built-in authentication with role- and attribute-based access let browsers and phones query it directly.

Every query was a transaction. Read-write queries were strictly serializable and read-only ones serializable, across regions, without relying on synchronized clocks. The protocol followed Calvin, a research system published at SIGMOD in 2012: a coordinator ran the transaction speculatively at a recent snapshot and buffered its writes, the transaction was then appended to a replicated log, which was the only step that needed consensus, and each replica re-checked the reads at that log position, committing the writes if nothing had changed or aborting and retrying if it had. Event streams and event feeds pushed or pulled changes to collections or single documents.

In March 2025 the company announced it would close the service, and it shut down on 30 May 2025; data not exported by then was deleted. It promised an open-source version of the core database, and in April 2025 published it as the fauna/faunadb repository under Apache 2.0, a Scala and JVM codebase built with sbt, together with its existing drivers and CLI. The FaunaDB site describes the project as at a very early stage with no official releases, and the repository has not changed since May 2025.