Supabase

Supabase gives Pensieve an understanding of your project's database: which tables exist, what each one holds, how they relate, which parts of the organisation they serve, and how to query them. Nothing from the database is copied into your map — the connector is a routing map, not a mirror.

What you choose

Pick the schemas to describe. Your own schemas are listed first; Supabase's platform schemas (auth, storage and the rest) are listed too, flagged, so choosing one is deliberate rather than a default.

Each selected schema becomes a branch of one database tree — the same kind of tree a GitHub repository produces — with a Page for each table or cluster of tables describing its columns, keys, row-level security and the questions it answers. The tree lives beside your organisation's map and links into it, so a project or system Page can point at the data behind it.

Authentication

Sign in to Supabase and choose the project. You'll enter the project reference first (Project Settings → General) — it scopes everything Pensieve reads. Every read then goes through Supabase's read-only query endpoint, so Pensieve cannot write to your database even by mistake.

What lands in your map

Pages describing the selected schemas: tables, views, enums and functions; their structure, relationships and access rules; and how to query them with the Supabase MCP or SQL. While building those Pages Pensieve may read a handful of sample rows to understand a column, and never retains or reproduces them.

Keeping it current

Pensieve re-reads the schema daily and rebuilds only what changed. Saving a new selection, or pressing Sync, reads it straight away.

What it can't see

• Rows, beyond the small samples read while a Page is being written — and those are never kept.
• Schemas you did not select.
• A project the signed-in account cannot access.