The premise: for a vault, the license is a security property
Every other category on this site grades licenses as a rights question. For a password manager, the license is also a security question: open code can be audited, a self-hosted server removes the vendor's vault store as a breach target (the 2022 LastPass breach, in which encrypted vaults were exfiltrated, is the standing example), and a local file removes the server entirely. The method follows this site's open source discipline from the IFTTT and Jira benches: license read per project, the open-versus-closed boundary stated exactly, features third. The neighboring benches: business for admin buyers, free for tier audits, and the Bitwarden open source guide for the incumbent of this bench viewed from its own side.
| Project | License and boundary | Server model |
|---|---|---|
| Bitwarden | AGPL-3.0 core; some components under the Bitwarden License | Official server, cloud or self-hosted |
| KeePassXC | GPL | No server; a local encrypted file you sync yourself |
| Vaultwarden | Community implementation of the Bitwarden API | Self-hosted only; not Bitwarden's own code |
| Passbolt | AGPL-3.0 | Self-hosted or cloud, team-first design |
| Proton Pass | Clients open source; server closed | Vendor cloud only |
Grading the bench
Bitwarden is the center: its server and clients are predominantly AGPL-3.0, with specific components under the Bitwarden License, a split this guide states exactly because "Bitwarden is open source" as a slogan hides it. The practical meaning: you can self-host the official stack free for personal and organizational use, audit the code your vault depends on, and the licensing nuance matters mainly to those building commercial offerings on it. KeePassXC is the purity answer: GPL-licensed, local-first, your vault is a file (KDBX format, readable by a whole ecosystem of clients) that never requires anyone's server; sync is a file-sync problem you solve with tools you already trust, and there is no vendor to breach because there is no vendor.
Vaultwarden needs its honest paragraph: it is a community-written server, compatible with Bitwarden clients, far lighter to run than the official stack, and widely used. It is not Bitwarden's code and not audited to the same standard; choosing it is a deliberate trade of assurance for lightness, fine for a household, a real conversation for an organization. Passbolt is the team entry: AGPL-3.0, built around sharing credentials among colleagues with OpenPGP-based encryption, self-hostable, and the natural pick when the vault's job is team secrets rather than personal logins. Psono and gopass serve the edges (enterprise self-hosting and command-line git-native workflows respectively). Proton Pass is the boundary case that proves the method: its clients are open source, its server is not, so it is a privacy-brand cloud product, not an open source vault, and it is graded on the free bench rather than this one.
The comparison tool below puts Bitwarden, KeePassXC, and Vaultwarden side by side from the dataset; the checklist is the self-hosting audit, because an unpatched vault server you run yourself is worse than the cloud you left.
