BadgerDB is an embeddable, persistent, fast key-value database written in pure Go, aimed at Go developers who need local high-performance storage inside their own applications rather than a separate database server.

What it is

BadgerDB is a key-value store written entirely in Go and intended to be embedded directly in an application as a library. It is the underlying database for Dgraph, a fast distributed graph database, and the README describes it as a performant alternative to non-Go-based key-value stores such as RocksDB. The project is stable and, according to the README, is used to serve data sets worth hundreds of terabytes. It carries the Apache-2.0 licence, and its topic list places it among databases, document databases, key-value stores, Go libraries, and SSD-oriented storage.

The concrete problem it solves is the one a Go service hits when it needs durable local storage and the available engines sit outside its toolchain. RocksDB is not written in Go, and comparable embedded stores such as LMDB and BoltDB are also foreign to it. Badger keeps that layer inside Go, so it installs through Go modules and links into the same binary as the application. It handles the consistency work that would otherwise fall to the caller: concurrent ACID transactions with serializable snapshot isolation (SSI) guarantees, and persistence behaviour that has been checked against filesystem-level anomalies. Crash resilience is treated as a first-class concern, with the project publishing an ALICE write-up on that theme.

Key capabilities

• Embeddable pure-Go key-value library, consumed through go get github.com/dgraph-io/badger/v4.
• Concurrent ACID transactions with serializable snapshot isolation (SSI) guarantees.
• A Jepsen-style bank test that runs nightly for 8 hours with the --race flag to confirm transactional guarantees hold.
• Testing against filesystem-level anomalies to verify persistence and consistency.
• A badger command line tool, installed with go install . from the repository, which performs operations such as offline backup and restore.
• Built with Go 1.23, with v3 and above requiring Go modules; the version is deliberately held back to limit downstream effects on applications built with older Go releases.
• Suitability for SSD-backed storage, reflected in the project topics and in the SSD-oriented design of the engine.

Who uses it and how

• Dgraph uses Badger as its underlying database, which is the largest single proof of the engine running in a distributed graph product.
• Jaeger Tracing embeds Badger as its storage layer, a case where a Go service needs durable local state without deploying a separate database.
• UsenetExpress is listed among projects running Badger in production.
• Workloads reach hundreds of terabytes, so the engine is not confined to small local caches.
• Teams shipping Go programs can keep a single binary deployment shape, because Badger links into the application instead of running as a separate process.

Getting started

Install Go 1.23 or above, then run go get github.com/dgraph-io/badger/v4 from the project to retrieve the library. For the command line tool, check out the repository, run go install . inside the badger directory, and the utility lands in $GOBIN; documentation lives at badger.dgraph.io.

How it compares

The facts do not name any paid products that Badger replaces, so the comparison rests on the similar tools the README and blog posts do name. Badger positions itself specifically as the pure-Go alternative to RocksDB, and the project's own benchmarking material contrasts it with LMDB and BoltDB, both embedded key-value engines available to Go programs. All of these run locally and are licensed independently of Badger, which is Apache-2.0.

When to use it — and when not to

Choose Badger when a Go application needs an embedded, transactionally consistent key-value store with SSD storage and the team is willing to treat persistence as part of the application rather than as a separate service. Do not choose it if the requirement is a standalone server with a network protocol, multi-language clients, or operators who expect to manage a database independently of the code, since nothing in the README describes those capabilities. One honest gap in the supplied material: the Design section is truncated mid-sentence, so the engine's internal design goals and tuning guidance have to be read from the external documentation rather than from the README alone.

go get github.com/dgraph-io/badger/v4

cd badger
go install .