You wouldn’t let an agent write code if it wasn’t in Git. Why would you let an agent write to your database if it wasn’t in Dolt? We are Git for Data after all.
There’s a key difference between traditional databases and Dolt databases, and it’s fairly simple once you understand our goal: version control. Agents need it, and this video helps to paint that picture. We’ve also written a bunch of blogs about the thesis if you prefer a good read.
Traditional databases fall short in how agents interact with them. Multiversion concurrency control (MVCC) gives users or agents a very short rollback window, but miss the window, the data is gone after the change. Dolt gives you an infinite rollback window by swapping out B-trees for Prolly Trees, opening up a world of possibilities for agentic work.
Why Agents Need Version Control#
- Create instant branches, completely isolated from main, to test production data changes.
- A branch per agent means multiple agents can work on real data in parallel.
- Only changed data is stored.
- Every change is tracked, auditable, and revertible.
- Dolt’s built-in test engine means nothing gets to main if it doesn’t pass your tests.
- Rollbacks are instant.
Get Started#
Dolt Databases are free and open source and designed to be a drop-in replacement for your favorite DBs. Explore our current lineup here. If you want to learn more, visit our Docs, or ask us directly on our Discord.
PS, No AI was used in the making of this video. If you enjoy the video, please like or share it! Your support is always appreciated!