Braiden Blog
do not disturb

Why we're rebuilding LatchID from ground zero

LatchID runs on Authentik today. Here's why we're building our own auth engine to replace it, and how I'm steering the build.

If you scroll to the bottom of latchid.org, you'll see Powered by authentik. Authentik has carried LatchID so far, and it's been great. It gave us SSO, enrollment, recovery and OIDC/SAML without writing an identity provider from scratch.

But LatchID isn't a single login page anymore. It's turning into a platform with free personal accounts, paying teams, branded subdomains, custom domains, and staff who need to manage all of it. So we're doing the scary thing: building our own auth engine from the ground up. I'm directing the build: setting what it needs to do, making the calls on scope and architecture, and reviewing what lands. A small team of developers is writing the code.

Where Authentik stops fitting

Authentik is an identity provider. LatchID is becoming a multi-tenant identity platform, and the gap keeps showing up as glue code.

  • Multi-tenancy is bolted on. Every Flagship brand (brand.latchid.org) is really a pile of flows, stages and naming conventions held together by discipline.
  • Business logic lives in a middleman. Self-service for teams works by putting an API in front of Authentik that works out the brand and role from an app's slug prefix. It works, but it's a second system pretending to be one.
  • Staff access is all-or-nothing. I want LatchID staff to see what their team and role allow, across every tenant, not just "admin" or "not admin".
  • Plans aren't a concept it knows. Core, Flagship, Flagship+ and BYO have real differences in app ceilings, LDAP, custom domains and audit logs. Those end up enforced around the IdP, not inside it.

None of this is Authentik's fault. It's just not what it was built for.

What the new engine does

It's one backend that runs everything hitting auth.latchid.org and every brand's login page, with different frontends on top.

The baseline is full parity with everything we use Authentik for today. On top of that, it adds what we actually need:

  • Real multi-tenancy where tenants rope back to main auth. If your org pays for Flagship, you can sign into its apps with your work email from your normal Core account.
  • Team- and role-based permissions for staff, everywhere
  • Plan limits enforced by the engine itself, which BYO plans need most since every one is a different mix
  • Spinning up a new node when we need one, without a weekend of hand-config

How it's laid out

It started as five repos. That was too many. Now it's three:

RepoWhat it does
coreThe engine: every login, every brand, every token
frontendThe main latchid.org site
extra-nodeHow to provision a new node

Not a weekend project

This is a long build over an extended stretch. The developers do the heavy lifting on the code; my job is keeping everyone pointed at the same target and making sure what ships is what LatchID actually needs. Authentik stays in production until the new engine can fully take its place. No big-bang cutover.

Why bother?

Because what I'm really selling is trust in a login. If the most important part of LatchID is someone else's software plus our glue, we don't fully own it. A custom engine means every flow, permission and tenant boundary is something we designed on purpose, and something I can answer for end to end.

More as it comes together.