DoltHub

Market position

The company describes itself as a database company, having shifted from its original data sharing/marketplace positioning after prospective customers asked to run Dolt as a production OLTP database. It presents Dolt as the world's first and only version-controlled SQL database and positions it against MySQL for compatibility and performance comparisons, with Doltgres targeting PostgreSQL users, DoltLite targeting SQLite and Dumbo targeting MongoDB. DoltHub is positioned as a GitHub-equivalent for Dolt databases.

The company describes Dolt as the first and only version-controlled SQL database, combining full Git-style branching, merging and history with MySQL wire compatibility, so that the same database can be operated with either SQL clients or Git-equivalent CLI commands. Its storage relies on prolly trees, a data structure pioneered in the Noms project that enables structural sharing and efficient storage of many database snapshots along with fast diffs. Dolt can also be attached to an existing MySQL deployment as a versioned replica, avoiding a migration while adding per-cell audit history, time travel and branch-based development.

Technology

Dolt is written in Go with a cgo dependency, distributed as a single ~103 MB binary, and licensed Apache-2.0 on GitHub. Its versioning is built on prolly trees, the data structure first implemented in the Noms project, which enable structural sharing, fast diffs and history independence. The SQL layer is built on go-mysql-server, a storage-agnostic MySQL-compatible query engine in Go that the company adopted, forked and now maintains as the authoritative project after src-d ceased operations in late 2019. A new journaled storage format launched in August 2022 and completed in March 2023 delivered ACID compliance and improved performance. Related codebases include DoltgreSQL (PostgreSQL-flavored, in Go), DoltLite (in C), DumboDB (document database, in Go), a Go database/sql driver, and the TypeScript Dolt Workbench. The DoltHub web stack reportedly includes Node.js, Apollo, Tailwind CSS, Envoy, Emotion, Ace, Stripe, Google Sign-in, HubSpot Analytics and Google Analytics.

Go-to-market

Distribution is led by open source: the Dolt repository has over 24,000 GitHub stars and is packaged through Homebrew, MacPorts, Arch Linux, Chocolatey, a Windows installer and Docker images, with a community Discord and a public roadmap. Demand generation runs through a high-frequency engineering blog and a weekly mailing list of updates (archived publicly since July 2022), plus free hosting of public databases on DoltHub as an on-ramp. Commercial conversion happens through paid private repositories, support contracts, the DoltLab on-premises product for organizations unwilling to host data publicly, and the Hosted Dolt managed service. Inbound demand from prospective customers wanting to run Dolt as a production OLTP database, rather than for data sharing, drove the company's shift in product focus.

Developers and engineering teams that need version control over database contents, including teams running Dolt as a production OLTP database (an early example being Turbine), organizations wanting DoltHub-style collaboration inside their own network via DoltLab, teams that want a managed Dolt server through Hosted Dolt, users of public shared datasets on DoltHub, and, more recently, builders of AI agent systems needing agent memory. A federal SBIR Phase I award for a project titled "Evolving the database for the USAF" indicates U.S. Air Force interest.

Geography

Headquartered in Santa Monica, California, United States, with addresses listed at 1420 2nd St, Santa Monica, CA 90401 and 1230 Georgina Avenue, Santa Monica, CA 90402.

History

The founder traces the idea to 2013 ("data as a service") while at Amazon Web Services. Noms, the technical predecessor that pioneered the prolly tree data structure, received its first commit on 2015-06-02 by Aaron Boodman and launched publicly in August 2016; Attic Labs, the company behind Noms, was acquired by Salesforce in January 2018. Liquidata was incorporated in February 2018, and the three founders — who had quit jobs at Snap — began working full time on 2018-08-06. In October 2018 a Git-style command line was added on top of Noms, and in November 2018 the resulting tool was named Dolt; work on DoltHub began in December 2018. Dolt launched on 2019-08-06 and DoltHub in September 2019. The company rebranded from Liquidata to DoltHub in September 2020. Data Bounties launched in December 2020 and ended in September 2023. The first production OLTP customer, Turbine, began testing in January 2021, and the company announced OLTP intentions in June 2021. DoltLab launched in January 2022, Hosted Dolt in May 2022, a new storage format in August 2022 (completed March 2023), Dolt 1.0 in May 2023, and Doltgres in November 2023. In 2026 the product line expanded further with Doltgres 1.0, DoltLite and DumboDB.

Risks & controversies

The company has publicly documented that its original data marketplace and data sharing thesis did not materialize: the Data Bounties program, launched in December 2020 as an attempt to bootstrap a data sharing ecosystem, was ended in September 2023 so the company could focus on being a database company. Early technical constraints included a storage format inherited from Noms that was a performance bottleneck and required migrating old-format databases, and an early production customer tolerating an approximately 20x performance penalty relative to MySQL that the company initially could not meet on some queries. Adoption of a database at the storage layer also depends on customer trust, which the company identified as a major internal theme.