Open-source · Deploy anywhere · Postgres-nativePoint it at the database you already run. Nothing to provision, nothing copied, nothing migrated — your schema is the only input.
01·Code as Truth
One TypeScript file per collection is the entire input. The database schema, the REST API, the typed SDK, the forms and every admin view come out of it — and change with it.
import type { PostgresCollectionConfig } from "@rebasepro/types"; export const products: PostgresCollectionConfig = { name: "Products", slug: "products", table: "products", properties: { name: { name: "Name", type: "string", validation: { required: true }, }, category: { name: "Category", type: "string", enum: { electronics: "Electronics", fashion: "Fashion", home: "Home & Garden", }, }, price: { name: "Price", type: "number" }, in_stock: { name: "In Stock", type: "boolean" }, image_url: { name: "Image", type: "string", url: true }, }, };
→ generates Admin views · forms · DB schema · REST API · Typed SDK
02·What your app talks to
A typed SDK for your collections, REST over every table, and realtime on a WebSocket — plus auth, storage and backups, running against your database from minute one.
Plus auth and roles, S3-compatible storage, scheduled jobs, serverless functions and backups — all against the database you already own, under the same policies.
Explore the backend03·Security-first
Row-level security written in the same TypeScript file as the collection, compiled into real Postgres policies. Granular per collection, per field, per role — and enforced by the database itself, not by middleware someone has to remember to call.
rebase.uid() = idrebase.uid() = idrebase.uid() = idauth.role() = 'admin'rls-check reads your database's own catalogue and reports what is actually exposed — tables served with row-level security switched off, policies that evaluate to true for everyone, views that read straight past the RLS on their base tables. Fourteen checks. Any Postgres: Supabase, Neon, RDS, your own server. It is read-only, it needs no account, and it does not care whether you use Rebase.
npx @rebasepro/rls-check $DATABASE_URLSELECTs against the system catalogues and nothing else. No telemetry, no upload, and it never reads your data.
04·One definition, either half
The panel reads your data through the same API and the same policies your app does. Ship headless today and add it the day a human needs to touch the data — or never. Whichever half you leave out, the API answers exactly the same.
What your app talks to
REST, auth, storage, realtime and backups over your database. No UI, no React anywhere in the dependency tree.
@rebasepro/server
Where your team works
Spreadsheet editing, media, relations, roles and workflows — with your own React components dropped in wherever the default field isn't enough.
@rebasepro/cms
Developer tools, inside the panel
SQL editor, schema visualizer, RLS policy editor, logs and API explorer — registered as extra views inside the panel.
@rebasepro/app
05·Your workflow
The same Postgres rows as a data grid, cards, a list or a Kanban board. Switch layouts on the fly, filter live, edit inline — without building a single admin screen.
06·Who touches it
Writes one collection file and gets a schema, REST, an SDK and policies. Mostly stays in the editor — and when a question is faster to answer in SQL, opens Studio rather than a second tool.
The backend StudioFixes a wrong address, refunds an order, moves a card. Did not choose Rebase and will not read its docs — the panel answers to them.
The admin panelReads and writes through the same API under the same policies as the humans. A first-class caller, not an integration.
AI & agents07·Built for the agent era
An AI agent can scaffold a backend in an hour. It can't tell you whether that backend is safe. Agents are extremely good at producing plausible backends — and famously bad at producing secure ones. Rebase makes the safe outcome the only outcome, by construction.
Explore all AI & Agent featuresNative pgvector with cosine, L2 and inner-product distance. Query embeddings straight over REST — no separate vector database to run.
08·Built for real products
From marketplaces to retail loyalty, public data and internal tools — seven very different products, the same definition underneath.
Vector Design Tool · Built by the Rebase team
DadakiOur own product: a fast, in-browser vector editor with cloud sync. Dadaki runs on a completely standard Rebase backend — the same auth, storage, and generated APIs every project gets.
dadaki.com
Digital Pain Therapy
A certified digital medical device delivering AI-personalized pain therapy. Rebase runs the entire backend — patient records, therapy plans, the exercise content library, and the clinical dashboards on top of them.
medicalmotion.com
Marketplace & Job Portal
A two-sided marketplace connecting sustainability professionals with green companies. Rebase powers both sides of the market: role-based auth, typed matching APIs, and the moderation back office.
sustentalent.com
Public data · Built by the Rebase team
A dónde va el dinero de EspañaA public reading of Spanish government spending, from €725bn down to a single budget line. There is no database in the read path: a cron recompiles the official figures each month, storage serves them as JSON, and the page itself makes zero API calls.
presupuestos.rebase.website
AI sales agent · Built by the Rebase team
An outbound agent that reads your site, plans audiences, then finds real companies and the people at them — each one verified against its own website. Crons, functions and the review board are stock Rebase. It drafts; the schema has no state in which it sends.
prospector.rebase.website
Retail loyalty
The loyalty programme behind an apparel brand. Orders from Odoo POS and Shopify land in Postgres, a tier engine recomputes points and discounts, and the Apple and Google Wallet passes customers scan at the till are issued from the same schema.
unfeignedgear.comInternal tool · CRM and workflow automation
A CRM the team can rebuild without a deploy: pipelines, tags, tasks and an automation canvas whose triggers, branches and actions are all rows in Postgres. Twenty-three collections generated the entire back office — no admin screens were written.
Internal tool — no public site
01/07
The changelog is the record. Here is what is being built next.
One command against the Postgres you already have. No account, no container to pull, nothing to sign up for.
pnpm dlx @rebasepro/cli initDocker, Fly, Railway, Hetzner or bare metal. Your data stays in your infrastructure and never reaches us.
docker compose up -dManaged hosting. We run our own products on it first — early access goes out from the waitlist, oldest first.
