One account per app
Every app ends up in its own account. Separate logins, separate keys, separate dashboards to keep straight.
Every project gets its own isolated stack in the EU. Keep supabase-js — change the URL and keys. Nothing else.
import { createClient } from '@supabase/supabase-js'
export const supabase = createClient(
process.env.SUPABASE_URL, // your Basequel URL
process.env.SUPABASE_ANON_KEY // your Basequel key
)
Runs in production for
Supabase is a brilliant way to start. Run a portfolio of apps on it — with agents doing real work — and the seams show.
Every app ends up in its own account. Separate logins, separate keys, separate dashboards to keep straight.
Connect an agent and it only reaches the account it signed in to. The rest of your projects are invisible to it.
Hit a usage quota and your project slows down. It rarely happens at a convenient moment.
Basequel puts every project in one place, each on its own isolated stack. One agent connection reaches all of them — within the scopes you grant.
Get early accessThe same building blocks and the same client libraries. Each project runs as its own stack, so one busy app never shares a database with another.
Postgres, Auth, APIs, Storage, Realtime, Edge Functions and Studio — provisioned per project, not shared. Point supabase-js at the new URL and keys and your app carries on.
A full, current Postgres. Your schema, extensions and SQL, as they are.
Email, Google and GitHub sign-in, compatible with the Supabase Auth client.
APIs generated from your schema automatically. No server code to write.
Buckets for files and images, with the policies you already use.
Subscribe to database changes and broadcast to connected clients.
Run server-side logic next to your data, migrated straight across.
The familiar dashboard for tables, SQL, auth users and storage.
Using Drizzle, Prisma or pg from a serverless function? Switch on a public, TLS-encrypted Postgres endpoint for that project only. Off unless you need it.
Everything comes across. Your users don’t notice.
One command copies schema, data, Auth users, Storage files, scheduled jobs and Edge Functions.
Point a branch or staging build at the new URL and keys. Passwords keep working. Scheduled jobs arrive paused, so nothing runs twice.
Re-copy right before you switch to pick up the latest data. URLs and keys stay the same, so there’s nothing to redeploy twice.
Shown for illustration. Migration tooling is available to early-access projects.
Give Claude, Hermes or your own agent one connection to every project — with guard rails you set.
Works as a claude.ai connector over OAuth, and with Claude Code, Hermes and other agents via tokens.
basql CLIThe same capabilities from a terminal or a script, for humans and agents alike.
Tokens are limited to the projects and actions you choose. SQL is read-only unless you say otherwise.
See which agent did what, when, and under which token.
orders and coupons. I can draft policies, but applying them needs a token with write scope for shop.| Agent | Tool | Scope |
|---|---|---|
| claude.ai | run_sql | shop · read |
| claude.ai | describe_schema | shop · read |
| claude.ai | list_projects | all · read |
Agent access is in active development. Tool names are illustrative.
A backup you’ve never restored is a hope, not a plan. Basequel restores one automatically every week.
Germany
Hosted with Hetzner in Nuremberg. Your data stays in the EU.
Strict TLS
Every request goes through Cloudflare with strict TLS, end to end.
One stack per project
No shared database between projects. One app can’t starve another.
Daily, encrypted
Encrypted daily backups to off-site object storage on Cloudflare R2, kept for 30 days.
Weekly restore test
An automatic restore every week proves the backups actually come back.
Scoped and audited
Read-only SQL by default, tokens limited to what you grant, every call logged.
Each backup is kept for 30 days. Days shown are illustrative of the weekly cadence.
Anything else? Mention it in the waitlist form and we’ll reply.
No. Basequel is Supabase-compatible, so your app keeps using supabase-js. Only the project URL and keys change.
Schema, data, Auth users, Storage files, scheduled jobs and Edge Functions — in one command.
Auth users keep their passwords, so nobody has to reset anything. Scheduled jobs are copied paused, so nothing runs twice while you test.
Re-copy right before you switch. The re-copy picks up the latest data and keeps your Basequel URLs and keys the same, so the build you tested is the build you ship.
In Germany, on Hetzner infrastructure in Nuremberg. All traffic goes through Cloudflare with strict TLS.
Every project is backed up daily, encrypted, to off-site object storage on Cloudflare R2, with 30-day retention. An automatic restore test runs every week to prove the backups restore.
Yes. Each project can switch on an optional public, TLS-encrypted Postgres endpoint for serverless apps that talk to Postgres directly. It’s off unless you need it.
Basequel’s MCP server works as a claude.ai connector via OAuth, and with Claude Code, Hermes and other agents via tokens. There’s also a CLI, basql. Tokens are scoped, SQL is read-only by default and every agent call is written to an audit log. Agent access is being built now and will reach early-access projects first.
Yes. It runs in production today for Prospel, AutoGenShorts, Intelligo, LatticeRun, GenKid and Selfscript — products from Sequel AI, the team behind Basequel.
Basequel is in early access. There’s no self-serve sign-up or billing yet. Join the waitlist and we’ll talk you through it when your project is onboarded.
We’re onboarding projects in small batches. Tell us a little about yours and we’ll be in touch.
supabase-js