Tables & schema alterations

Method	Route	Description
GET	/projects/{id}/tables	List tables, with their columns
POST	/projects/{id}/tables	Create a table ({ name, columns: [...] })
PATCH	/projects/{id}/tables/{table_id}	Rename a table ({ name }) — renames the physical table too
DELETE	/projects/{id}/tables/{table_id}	Drop a table
POST	/projects/{id}/tables/{table_id}/columns	Add a column ({ name, type, nullable?, default_value? })
PATCH	/projects/{id}/tables/{table_id}/columns/{column_id}	Rename a column and/or change its type/nullable/default (any subset)
DELETE	/projects/{id}/tables/{table_id}/columns/{column_id}?confirm=true	Drop a column — irreversible, so ?confirm=true is required

Column type is one of text, integer, decimal, boolean, timestamp, json. A column can never be named id, which is always the auto-incrementing primary key.

Running raw SQL

Method	Route	Description
POST	/projects/{id}/sql	Run arbitrary SQL against the project's tables ({ query })

This runs whatever SQL you send directly against the underlying database, authenticated with your platform token — there is no query sanitization, statement whitelist, or RLS applied at this layer, since it's meant as a platform-owner escape hatch (the dashboard's SQL runner uses this endpoint). Treat the platform token like a database credential: anyone with it can read, write, or drop anything in the project.