Many isolated stacks on one box
Self-hosted Studio is hardwired to /project/default. Tensorbase runs a full stack per project, each with its own Postgres, secrets and subdomain.
Supabase self-hosting gives you one project per install and no way to make another. Tensorbase is the control plane that was missing — many isolated projects on hardware you own.
$ tensorbase new goodsoil==> creating goodsoil (index 1)==> generating secrets api https://goodsoil.base.tensorcode.co.id studio https://goodsoil-studio.base.tensorcode.co.id proxy ip 10.10.1.41 pooler 127.0.0.1:5501 session / 6501 transaction $ tensorbase start goodsoil✓ 13/13 services healthy · 2.16 GBThe open-source repo ships the database and every service around it. The layer that manages projects is the product Supabase sells. Tensorbase fills exactly that gap and nothing else.
Self-hosted Studio is hardwired to /project/default. Tensorbase runs a full stack per project, each with its own Postgres, secrets and subdomain.
The control plane that creates projects is the part Supabase keeps closed. Tensorbase allocates the IPs, ports and secrets and brings the stack up.
Supabase's compose is vendored byte-identical and layered with one override, so an upstream release is a pull, not a merge conflict.
No second reverse proxy fighting for :80. Projects register as Traefik labels on your existing instance, with certificates over DNS-01.
The next free index derives a proxy IP, a database IP and loopback ports for the pooler. Nothing collides with a project you already run.
A fresh JWT secret, the anon and service_role tokens signed with it, Postgres and dashboard passwords. No default value ever reaches production.
Traefik picks up the labels, issues a certificate over the DNS-01 challenge, and the project answers on its own subdomain.
Not a trimmed build. The same thirteen services the hosted platform runs, isolated per project.