Every IdP sells one place to manage access, and Okta and Entra already offer passkeys. The difference is what the central directory is made of.
∅
No credential database to breach
The sign-in service holds nothing that proves who someone is. A stolen database or server can’t produce a login that a self-checking app will accept.
Postgres holds protocol state only: clients, codes, refresh families, SCIM cursors. Losing it logs people out; it doesn’t lose who works there.
⛓
A directory that can’t be quietly edited
Every hire, role change, and departure is a signed entry that needs several admins to approve. Auditors read the log instead of trusting screenshots.
Membership, roles, and device approvals live in the org provenance log. An m-of-n lock script enforces the admin quorum, so a compromised host cannot add itself an admin.
⚡
Leavers lose access in seconds
When someone leaves, one signed entry turns them off, and every BetterId sign-in checks it within seconds. Sessions are kept short, so there’s little left to chase.
The org node streams new log heads to validators, and the sign-in service refuses the next sign-in and revokes refresh tokens. CAEP session-revoked and SCIM-out deprovisioning are designed for later.
☎
No more “reset my password”
There’s no password, so there’s no reset, and no help-desk call for an attacker to talk their way through. A lost phone is handled by identity proofing and admin approval.
Recovery is an org-quorum key rotation on the employee’s work VLAD after policy-selected identity proofing. It is a signed, visible log entry.
◇
One identity for every door
The same work identity is designed to sign people into web apps first, then Windows and macOS, and later servers, VPN and Wi-Fi, when a customer needs them.
One work VLAD drives OIDC sign-in first, then PKINIT and Platform SSO; OpenSSH user certs, EAP-TLS and SPIFFE SVIDs from vlad-ca come later.
⇋
Partner access without federation projects
A partner’s staff prove who they are with their own company’s log. You decide which roles you accept; there’s no per-partner setup project.
/partners/<org-vlad> in your org log replaces per-pair SAML federation; partner membership is checked against their org log.
⌂
You run it, and it outlives us
The service and the network it depends on run in your environment. Sign-in keeps working if the internet’s public network, or Cryptid, is down.
The org node pins your org and employee logs, optionally on a private network. Logs verify without any Cryptid service.
↔
Built for the stack you have
The first release is a small sign-in service that sits behind the IdP you already run. Full editions in the languages your team uses come later, when customers ask, and share one set of security-critical code.
First, bs-orgd: a small OIDC service over one Rust core, plugged into Entra (then AD FS) as a factor. Java, .NET and Python editions come later, on request, gated by a shared conformance suite. See BetterId IdP.