Dolt is an open-source database that brings the full Git versioning model to structured data. Under the hood it implements a novel storage engine built on prolly trees that makes branch, merge, diff, blame, and log operations first-class SQL operations rather than external tooling. The MySQL wire protocol compatibility means any application that connects to MySQL can switch to Dolt without code changes, gaining automatic version history on every table.
The versioning primitives map directly to Git concepts. Developers create branches to experiment with data transformations, diff branches to see exactly which rows changed, merge branches with automatic three-way conflict resolution, and tag releases for reproducibility. The system call interface exposes these operations as stored procedures, so version control happens inside SQL transactions rather than through a separate CLI. DoltHub provides a hosted collaboration platform where teams can fork, clone, and submit pull requests on databases.
Dolt has found strong adoption in AI and ML workflows where training data lineage and reproducibility are critical. Teams use branches to A/B test dataset modifications, diff to audit data quality changes, and the full commit history as an audit trail. The database is also emerging as a persistent memory backend for AI agents, where each agent session can branch, write, and merge back. Dolt is Apache 2.0 licensed with over 21,000 GitHub stars and a commercial hosted offering called Hosted Dolt.