AI Assistant
Query and update your data in plain language
Natural Language to SQL
Managed installs include an AI assistant that understands your schema, available in your Supasheet app at https://<project-ref>.supasheet.app/ai. When you connect a project, the platform indexes your tables, columns, and relationships so the assistant can translate questions into accurate SQL:
"Show me the ten customers with the highest order total this quarter"
becomes a SELECT you can run, refine, or paste straight into a chart or dashboard widget. The assistant keeps the conversation context, so follow-ups like "now only for Europe" refine the previous query.
Reads and Writes
- Reads — the assistant generates and runs
SELECTqueries and shows the results inline. - Writes — for
INSERT,UPDATE, orDELETE, the assistant first shows a preview of the rows that will be affected and asks you to confirm before anything executes. Confirmed mutations return the changed rows so you can verify the outcome.
Everything runs as you: queries execute under your Supasheet role, so row level security and permissions apply exactly as they do in the rest of the app.
Guardrails
The assistant will not run DDL (schema changes) and refuses unscoped destructive statements — an UPDATE or DELETE without a WHERE clause is rejected rather than guessed at. Schema changes belong in Database Management, or let a coding agent build them through the MCP server.
Good Uses
- Drafting the SQL for charts, widgets, and reports without remembering exact column names.
- Ad-hoc questions during a meeting ("how many signups this week vs last?").
- Small, controlled data fixes with a visible preview before committing.
For anything the assistant gets wrong, the generated SQL is always visible — edit it by hand in the SQL editor.