Start

What dabadata is

dabadata hosts your whole app: its frontend on your own domain, a realtime database, sign-in for your users, file storage and AI calls, all under the same rules. Underneath it is plain Postgres, so your rows sit in ordinary tables you can open with any SQL tool. On top you declare your tables in one file, and rules that read like membership - "admins and editors may create, the owner may update" - and every decision comes back with a sentence that explains it. Every interaction is written to a trace, live subscriptions stream changes over one socket, and an MCP server lets an AI agent use the same data under the same rules.

Every feature is listed on one page, Features. AI assistants: the docs are indexed as markdown at /llms.txt, and every page has a markdown copy at its index.md.

How it fits together

1. Declare. One TypeScript file, projects/<name>.ts, lists the tables, their columns and one rule per operation. It is plain data - no functions hide in it. Tables, Rules.
2. Apply. On start, and whenever you call POST /v1/schema/reload, the server creates a Postgres schema for the project plus any missing tables and columns. It never drops anything or changes a type; it tells you about drift instead.
3. Hand out tokens. Your own backend holds the project key and mints a user token (id, name, role) for each signed-in person, your identity provider's tokens are trusted as they are, or dabadata hosts your users' sign-in itself. The browser only ever sees a user token. Identity and tokens.
4. Read, write, subscribe. The app calls /v1/mutate and /v1/db, or uses the JavaScript client, and keeps lists fresh over the live socket. Each call is checked against the rules and recorded in the trace with its decision and the reason.
5. Look inside. The console shows your projects, keys, users, team and usage, and the viewer the tables, the trace and a rules simulator; the MCP server lets an agent read the schema, change rows and explain a denial from the trace.
6. Put it online. A site serves your app's frontend next to its data, at <name>.dabadata.site on every plan or your own domain from Starter, deployed with one command or by your AI agent.

What is in the repository

Where to go next