Git for data - version control for databases with SQL interface
Dolt is a version-controlled SQL database that combines Git-style versioning with relational database functionality.
AI Panel Score
6 AI reviews
Reviewed
AI Editor ApprovedApproved and published by our AI Editor-in-Chief after full panel analysis.Dolt is a version-controlled SQL database that brings Git-like functionality to relational data management. It allows users to create branches, track changes, and merge database schemas and data with full version history, making it possible to collaborate on databases the same way developers collaborate on code.
The platform provides a MySQL-compatible SQL interface, enabling existing applications and tools to work with Dolt databases without modification. Users can perform standard database operations while automatically maintaining complete audit trails of all changes, including who made changes, when they were made, and what exactly was modified.
Dolt is designed for data teams, analysts, and developers who need to manage evolving datasets with accountability and collaboration features. It supports data sharing through DoltHub, a platform for hosting and discovering public datasets, and enables teams to work on the same database simultaneously through branching and merging workflows.
The system addresses challenges in data management where traditional databases lack proper version control, making it difficult to track data lineage, collaborate safely, or roll back problematic changes. Dolt combines the reliability of SQL databases with the collaborative power of Git version control.
Dolt provides cell-wise diffs via dolt_diff_<tablename> and dolt_history_<tablename> system tables, giving a built-in, native audit trail of exactly what changed in every cell, when, and by whom.
DoltHub and DoltLab support continuous integration testing workflows that allow teams to automatically validate data changes before they are committed to the primary branch.
Dolt supports long-lived database branches for write isolation and true SQL merges with conflict resolution, making it the only SQL database that can merge divergent data branches.
DoltHub is a web-based platform, similar to GitHub, where teams can share, discover, clone, and collaboratively build Dolt databases, with public databases hosted for free.
Changes made on a feature branch can be submitted as a pull request, allowing teammates to review diffs and discuss proposed data modifications before merging them into the main branch.
Every data change is recorded as a Dolt commit with a hash, author, timestamp, and message, enabling time-travel queries to inspect or restore the database at any prior point in time.
Dolt lets you fork, clone, branch, merge, push, and pull a SQL database exactly like a Git repository, with all Git commands mapped to their Dolt equivalents targeting table rows instead of files.
Hosted Dolt is a fully managed, cloud-deployed Dolt database service where users choose server size and replica count, and DoltHub provisions and operates the infrastructure, including backups and monitoring.
Dolt ships with a built-in MySQL-compatible server on port 3306, allowing any existing MySQL client, connector, or application to connect and run SQL queries without modification.
All Git-style version control operations (add, commit, diff, log, merge) are exposed directly in SQL as dolt_<command> stored procedures and system tables, requiring no CLI or special tooling.
Dolt can be configured as a MySQL replica so that teams gain version control features on top of an existing MySQL primary database without migrating or changing their production setup.
DoltLab provides the same web UI as DoltHub but runs entirely on-premises via Docker Compose on a single Linux host, so organizational data never leaves the company's own infrastructure.
Self-hosted, open-source version controlled SQL database for anyone to run on their own hardware
Free remote hosting for public Dolt databases; individuals and open data projects
For individuals or teams needing private or larger database storage on DoltHub
Managed, cloud-hosted Dolt database for developers wanting a low-cost trial instance
Production-grade managed Dolt database; pricing based on instance type and storage selected
Self-hosted DoltHub for small teams wanting on-premises data collaboration
Enterprise on-premises DoltHub for large organizations needing security, scalability, and support
Paid support contract covering Dolt, DoltHub, DoltLab, and all related products
Dolt is a genuinely unique versioned SQL database, but the vendor is small and lightly funded.
“Dolt is the only SQL database that branches and merges data, which is real differentiation. The catch is a small vendor with roughly $21 million raised and no obvious second source.”
DoltHub Inc is eight years old, founded 2018 in Santa Monica, and has raised about $21 million across two rounds, the last a Series B in 2021. That is a small war chest for an infrastructure vendor, and a board will ask about runway before it asks about features.
What you are buying is real, though. Dolt is the only SQL database that can branch and merge divergent data, and the dolt_diff system tables give you a native cell-level audit trail. PlanetScale runs a polished MySQL platform, but it does not merge data the way Dolt does. This is a genuine category of one.
However, that uniqueness cuts both ways. There is no second vendor if DoltHub stalls. Pricing is friendly: free open source, then $50 a month on DoltHub Pro past 1GB. Run a 90-day pilot on one non-critical dataset, confirm the open-source license is your fallback, then decide.
Being the only SQL database that merges data moves a team forward, though few peers have standardized on it yet.
Apache-licensed open source plus a self-hosted DoltLab option means a board sees a defensible, not sketchy, choice.
A MySQL-compatible interface on port 3306 lets existing clients connect without rewrites, so payback comes fast.
Branching, merging, and a native audit trail genuinely advance how a team manages evolving data rather than just cutting cost.
Eight years in market is reassuring, but roughly $21 million raised and a small team make this a runway question.
Data teams who need real version control and audit trails on evolving datasets.
Organizations who need a large, deeply funded database vendor for mission-critical systems.
Dolt makes a Merkle-DAG storage engine the substrate for treating data exactly like code.
“Dolt builds branch, merge, and diff into the storage layer rather than stapling them on as triggers. The architecture is a sound versioned-data bet, but Merkle-tree storage carries read overhead a plain primary doesn't.”
Dolt's whole pitch hangs on one structural decision: data isn't stored in B-trees but in a Merkle DAG, the same content-addressed structure Git uses. For a CTO weighing a versioned data substrate through 2029, that's what makes branch, merge, and Cell-Level Diff actual primitives rather than triggers stapled onto a normal MySQL.
The craft ceiling is real. The dolt_diff_<tablename> system tables and time-travel queries are exposed straight through SQL, so existing MySQL clients on port 3306 connect unchanged, and the engine has been generally available since 2019. Against Liquibase or Flyway, which version schema migrations only, Dolt versions the rows themselves.
But the catch is read performance. A Merkle-tree storage layer carries overhead a plain InnoDB primary doesn't, so Dolt fits audit-heavy and collaborative datasets rather than a hot transactional core. The Versioned MySQL Replica path lets you adopt it beside production instead of betting the OLTP path on it.
Dolt versions rows themselves where Liquibase and Flyway version only schema migrations, defining its own niche.
Branches, pull requests for data, and cell-level diffs match how engineers already reason about change.
A MySQL-compatible server on port 3306 and a Versioned MySQL Replica path let it slot beside existing stacks.
MySQL compatibility limits lock-in, but Merkle-tree storage means read overhead is a permanent design constraint.
Building version control into a Merkle-DAG storage engine is best-in-class craft, not a trigger-based shortcut.
Data teams who need a full audit trail and branch-merge workflows on evolving datasets.
Teams who need a high-throughput transactional database for a latency-sensitive production core.
DoltHub Pro just cut its sticker 10x to $5/month, and the open-source core costs nothing.
“The open-source Dolt engine is free to self-host, and DoltHub Pro now bills $5/month past the 100MB free tier. Hosted Dolt and a $5,000/month support floor are where real spend starts.”
Three things carry a price, and two are near-zero. The Dolt engine is open source, free on your own hardware. DoltHub Pro repriced in April 2026 — $50/month down to $5/month past a 100MB allowance, then $1/GB above 5GB.
The spend lives elsewhere. Hosted Dolt trial instances start at $50/month for 50GB, but production is instance-priced on AWS or GCP. The catch is the support tier: a Dolt Support contract opens at $5,000/month, negotiable. For a team of 50 running this in production, support is the real year-3 number, not the workbench.
ROI is unusually legible. The dolt_diff system tables give a native, queryable audit trail, so compliance value is measurable, not hand-wavy. Compare PlanetScale, which versions schema only — Dolt versions the rows too.
Free tiers and a credit-card $5/month plan ease procurement, though Hosted Dolt production needs cloud-instance modeling.
Open-source self-host carries no lock-in, and the $5/month Pro tier is month-to-month with no term commitment.
Every tier is published — open source free, DoltHub Pro at $5/month, Hosted Dolt trial at $50, no sales call to see them.
The dolt_diff and dolt_history system tables make audit and compliance value directly measurable in SQL.
Engine and Pro are near-zero, but Hosted Dolt is instance-priced and a support contract opens at $5,000/month.
Data teams who need a queryable audit trail without a per-seat bill.
Teams who want managed uptime but cannot budget a five-figure support contract.
Dolt gives data engineers real branch-and-merge for SQL, but raw query speed trails plain MySQL.
“Dolt makes a bad migration recoverable instead of a postmortem. But it pays for that with measurably slower reads on Sysbench.”
A data engineer judges a database by the migration script that quietly corrupted a table last Tuesday, not the demo. Dolt answers that directly. The dolt_history_<tablename> and dolt_diff_<tablename> system tables expose cell-level diffs in plain SQL, so you can query exactly which value changed, when, and by which commit hash instead of grepping backups.
The workflow fit is genuine. Long-lived branches give you write isolation, and every Git operation maps to a dolt_commit or dolt_merge stored procedure, so versioning works from any existing MySQL client on port 3306 without a separate CLI. Dolt can also run as a Versioned MySQL Replica, layering history onto a production primary without a migration.
The catch is throughput. Dolt 1.0, stable since May 2023, runs roughly 1.9X slower than MySQL on Sysbench, with reads hit hardest. Neon branches a Postgres database with copy-on-write and stays faster, but it cannot merge divergent data the way Dolt does.
A bad migration becomes a dolt_diff query and a revert, not a backup-restore postmortem.
Docs cover system tables and merge conflict resolution with the depth of engineers who run the tool.
The roughly 1.9X Sysbench slowdown is a friction point felt across every read-heavy working day.
Time-travel queries, long-lived branches, and CI testing scale well past the beginner clone-and-commit path.
MySQL-compatible server on port 3306 and dolt_ stored procedures fit existing clients without new tooling.
Data engineers who need auditable, mergeable changes to evolving datasets.
Teams running latency-sensitive production workloads on raw query speed.
Dolt makes a database behave like a Git repo, and most days that actually pays off.
“A SQL database where branch, merge, and diff work on rows instead of files. The catch is you have to learn version control habits before any of it feels natural.”
Most databases forget. You change a row, the old value is gone, and figuring out who touched it later means digging through logs that may not exist. Dolt remembers everything. Every change is a commit with a hash, an author, and a message, and the Cell-Level Diff shows you exactly what moved. That alone is worth a look.
The daily feel is familiar in a good way. It speaks MySQL on port 3306, so existing clients connect without fuss, and DoltHub gives you a GitHub-style web workbench with pull requests for data. First 1GB is free, then DoltHub Pro is $50 a month. PlanetScale also does database branching, but it branches schema, not the actual rows.
The catch is the learning curve. If branch and merge are not already muscle memory, day three is homework, not welcome. Launched in 2019, Dolt is steady, just not effortless on hour one.
Cell-Level Diff and a GitHub-style DoltHub workbench show real attention to the small repeated motions.
Powerful at month three once branching clicks, but discoverability in the first hour is thin.
A version-controlled SQL database is desktop and server work, so mobile is reasonably out of scope.
First 10 minutes assume you already know Git branch and merge, so it leans toward homework.
Every change is an immutable commit with hash, author, and timestamp, so nothing silently disappears.
Data teams who already think in Git branches and commits.
People who just want a plain database without version control overhead.
A genuinely novel database that has stayed niche for seven years and counting.
“Dolt has shipped daily since 2018 and reached a stable 1.0 in 2023. The catch is a small market that has never gone mainstream.”
"World's first version controlled database" is the kind of superlative I usually discount. This time the claim mostly holds. Nobody else merges divergent SQL data branches.
The engineering is real. Dolt has shipped on a near-daily cadence since 2018, hit a stable 1.0 in 2023, and exposes cell-level diffs through dolt_diff_ system tables. But it runs roughly 1.9X slower than MySQL on Sysbench, by the team's own published numbers. Against PlanetScale, which gives you MySQL branching without the speed tax, that gap matters for hot paths.
Exit is clean — it speaks MySQL wire protocol, so you migrate out with a standard dump. The yellow flag is reach. Only $21M raised, Series B, and seven years in, version-controlled databases are still a niche, not a category.
True SQL branch merging is a genuine gap no named competitor fills.
MySQL wire compatibility means migration out is a standard dump, not a rebuild.
Only $21M raised and a niche market leave the three-year bet uncertain.
The "world's first version controlled database" claim is largely defensible, not pure hype.
Near-daily shipping since 2018 and a stable 1.0 in 2023 match a durable-vendor pattern.
Data teams who need auditable branching and merging of datasets.
Teams who need raw query throughput at MySQL speed.
Common questions answered by our AI research team
Dolt offers a MySQL-compatible SQL interface, meaning standard MySQL queries work with Dolt databases.
Yes, Dolt supports branching and merging of database schema changes, similar to how code branches work in version control systems.
Dolt maintains a complete audit trail of all database modifications, giving teams full visibility into who changed what and when.
Dolt tracks both schema and data changes, providing version control across the full relational database — structure and content alike.
