Skip to main content
Database holds table collections — structured rows agents and workflows can read and write. Collections belong to the selected end user in the top bar. Switch the End user if the list looks empty. Open Database. You do not create collections from a form on this page. Create and fill them with database tools in chat or workflows (header action Chat with Tables).

List

What you see

  • Search: Search collections…
  • Refresh and Chat with Tables
  • Database storage panel (usage, almost full / full, get more storage)
  • Cards per collection: name, description (or placeholder), document count, View, Delete

Read-only share (MCP)

On each card, Read-only share with project members:
  • On — other project members can read this collection via MCP; they still cannot edit or delete it from your ownership rules
  • The collection remains in your end-user list in the dashboard

Empty state

No collections yet — create them with database tools in chat or workflows → Chat with Tables.

Delete

Deleting a collection permanently removes its documents. Confirm in the dialog.

Collection detail

Open a card (URL uses the collection id). Back · name · description · Chat with Document · Refresh

Collection information

  • Total documents
  • Field count when a schema exists
  • Schema badges (field: type, required markers)

Documents

  • “Showing X of Y documents”
  • Search documents
  • Add Row — disabled until a schema exists (“Define a schema…” via tools/chat first)
  • Grid columns from schema/documents; always includes _id and External user (_ownerEUID)
  • Inline cell edit; JSON view for complex values
  • Owner cell: You (owner) or pick another external user via Select external user…
  • New row: _id may show auto; set owner; Save / Cancel
  • Delete document with confirmation

Attach to an agent

In the workflow editor, open Data access:
  1. Under Tables (database collections), tick the collections this agent may use.
  2. Optionally configure Tables for end users presets (for example private per user, shared read-only, submit only) and advanced row/column permissions.
  3. Search / table tools for those collections are added automatically when the agent runs — you do not pin a separate search tool per collection for this wiring.
Empty message if none exist: create some from Database first. Only collections in the current project (for the relevant end user) appear. See Workflows in the dashboard.

Tips

  • Define clear field names and types so agents can use the schema reliably.
  • Keep operational data here; use Knowledge for document Q&A (RAG).
  • For API CRUD: Table. Native tool slugs: Native tools: Database.