What is Nitro Repo?

Nitro Repo is an artifact manager. It sits between the tools that produce build outputs — mvn deploy, npm publish — and the tools that consume them, storing what was published and serving it back with the metadata each ecosystem expects.

It is free, open source, and self-hosted. The backend is Rust; the frontend is Vue.

What an artifact manager is for

A build produces artifacts: a jar, a tarball, a POM, a set of checksums. Something has to hold them so the next build can find them. You could use a shared directory and a web server, and for one project that works. It stops working as soon as you need any of the following:

• Resolution metadata. Maven asks for maven-metadata.xml to answer “what is the newest version”, and for a per-version metadata file to resolve a -SNAPSHOT to an actual timestamped build. npm asks for a packument listing every version and its dist-tags. Neither is a file your build produced; both have to be generated from what was published.
• Access control. Some repositories are public, some are internal, and the client sends credentials in an ecosystem-specific way.
• Caching upstreams. Every build pulling straight from Maven Central is slow, fragile, and invisible.
• Knowing what is in there. Searching, browsing, and seeing when a version was published.

What Nitro Repo does today

Maven	Hosted and proxy repositories. Generated maven-metadata.xml, including timestamped snapshots. Checksums verified on upload and generated on demand. Push rules enforced.
npm	A hosted registry. Scoped packages, dist-tags, publish, unpublish, deprecate, search, and both login flows — including the browser one npm 9 and later default to.
Storage	The local filesystem, a self-contained single-file object format, or S3 and anything S3-compatible. Chosen per storage at runtime.
Search	A query language over projects and versions, exposed as both plain-text search and full query syntax.
Badges	Generated SVG badges for a repository or a project.

Repository types are a trait, not a hardcoded list, so adding one is a matter of implementing it rather than restructuring the server. NuGet, Cargo, Docker and apt/RPM are open issues, not shipped features — Compared with others has where things actually stand.

What it is not

• Not a build tool. It stores what your build produced; it does not produce it.
• Not clustered. One server, one Postgres. Storages can be S3, so the artifacts themselves can live anywhere, but there is no coordination between multiple Nitro Repo processes.
• Not stable yet. 2.0 is in beta and breaking changes are still on the table.

Why it exists

The alternatives are either expensive, heavy, or both. Nexus is clunky; JFrog costs money at a scale most people never reach. Nitro Repo aims at the case in between: a single team, a handful of repositories, a machine that already exists, and no appetite for running a JVM the size of the artifacts it serves.

Next

• Installation — Docker, a binary, or a build from source.
• First run — the administrator, a storage, a repository.
• Core concepts — how storages, repositories, projects and versions nest.