Your database.
Documented and controlled.
The Tables Viewer is where you document every table and column in your database — what it means, who can read it, and what AI models are allowed to do with it. That documentation flows directly into every AI conversation on the platform.
Why It Exists
Your data is in your database now. Make sure everyone — human and AI — knows what it means.
When you move data out of HubSpot and into your own Supabase database, you gain full control.
But a raw database full of tables like hubspot_contacts and columns like hs_lead_status means nothing without context.
The Tables Viewer is where you add that context. Comments you write here get embedded into the AI's understanding of your data — so when an AI writes a query or analyses your records, it knows what each field actually means in your business.
The Interface
Browse every table, expand every column.
Table list
Every table registered with the platform is shown in an expandable list. Search by name or filter to tables that are missing documentation.
Column schema
Expand any table to see every column: its data type, whether it can be null, and any comment you've added.
Real schema, not guesses
Column names and types are read directly from the live database schema — not from a cached list. What you see is what's actually there.
Access Control
Who can do what with each table.
Permissions are set per table and cover four operations: Read, Write, Delete, and Schema (view the structure). They're set separately for humans and AI models.
Who can query rows from this table.
Who can insert or update rows.
Who can delete rows.
Who can see the table structure (column names, types, comments).
Each permission can be set to all users, no users, or specific users (allowlist or denylist).
AI Model Access
The same four permissions apply to AI models — separately from humans.
Every AI conversation on the platform has access to your database — but only within the limits you set here. A model with Read access to a table can query it and show you results. Without Write permission, it cannot change anything, even if it tries.
This separation matters because the AI and your staff have different use cases. A finance table might be readable by the AI (for analysis) but writable only by admins.
Default starting position
New tables default to Read-only for models and Read/Write for authenticated users. Tighten or loosen as needed per table.
Documentation
Comments aren't just notes. They're AI instructions.
Every comment you add here is stored in the database and injected into the AI's context whenever it's working with that table. Good comments mean better AI output.
Table Comments
What is this table for?
Add a one or two sentence description of the table's purpose. Click the pencil icon next to any table name to edit. Saved immediately.
Example: "Contacts synced from HubSpot. One row per contact, updated hourly."
Column Comments
What does this field mean?
Expand a table, click the edit icon next to any column, and describe what it stores. For HubSpot-sourced fields, clarify the business meaning behind cryptic names.
Example: "hs_lead_status: HubSpot lifecycle stage. Values: New, Open, In Progress, Qualified, Unqualified."
Auto-suggest (AI)
Let the AI draft comments for you
For tables where the column names follow a clear convention (like HubSpot synced data), you can trigger an AI-assisted comment generation that pre-fills descriptions based on the column name and type. You review and save each one.