Skip to content

raia Command · Section

Tables

Import structured data, manage schemas and versions, and give agents precise read-only access.

Tables turn spreadsheet data into structured, searchable records that people can inspect and approved AI agents can query precisely. They preserve rows, column types, filters, totals, and stable references rather than treating the spreadsheet as a document.

In this section

  • Choose Tables for structured records such as invoices, inventories, transactions, contacts, and price lists.
  • Import and review CSV, TSV, or Excel data before publishing it.
  • Manage data, schemas, instructions, privacy review, re-imports, versions, and rollback.
  • Give external AI tools read-only access to selected Tables through a scoped Tables MCP link.
The Data Room three-dot menu with the Tables entry among Select root folders, Knowledge Graph, and Downloads
Tables open from a project's Data Room: choose the three-dot menu in the toolbar and select Tables to switch from documents to structured records.
The Tables page of a project's Data Room, with the Tables and MCP links tabs and the Import button
The Tables page holds your imported tables on the Tables tab, read-only MCP links on the MCP links tab, and the Import button at the top right.
01

What's inside

  1. 01

    Importing and reviewing a spreadsheet

    Open a Knowledge area, choose Tables, then Import table. Command accepts CSV, TSV, XLSX, XLSM, and XLS files up to 25 MB, 50,000 rows, and 200 columns. Choose one or more workbook sheets; each selected sheet becomes a separate Table. Macros are never run. Command proposes names, types, descriptions, identifiers, and cleaning rules for review, with a basic manual proposal available when AI analysis cannot run.

  2. 02

    Publishing and status

    Review the cleaned preview and turn suggested cleaning rules on or off before publishing. Column system names stay stable for agent queries and citations. Larger Tables can finish publishing in the background; the list shows whether each import is analyzing, waiting for review, publishing, ready, or failed.

  3. 03

    Data tab

    Browse 100 rows at a time, sort by column, and apply type-aware conditions including equals, contains, starts with, greater than, less than, one of, and is empty. Word search finds whole words across text columns. CSV export follows the active filters and word search; editors can select rows or matching results for deletion.

  4. 04

    Schema and agent guidance

    The Schema tab is the data dictionary: maintain display names, descriptions, data types, units, synonyms, and the identifier column, then download it as Markdown or JSON. In Settings, rename the display name without changing the stable system name and add Instructions for agents explaining what the Table contains and when it should be used.

  5. 05

    Visibility, privacy, and testing

    A Table must be Ready, marked Visible to agents, and granted to the exact Tables MCP link before an external agent can read it. Command reviews Table data for possible personal or sensitive information; high- or critical-severity findings keep it hidden until an editor reviews the masked finding and records an explicit acknowledgement. Test with an agent lets you inspect the request and returned rows before connecting another tool.

  6. 06

    Versions, re-imports, and rollback

    Re-import publishes updated source data as a new version after you review added, removed, renamed, or changed columns. Version history records row count, date, status, privacy-review status, size, and warnings. Rolling back restores the selected published version exactly, including rows or columns changed later.

  7. 07

    Tables MCP links

    Create a separate read-only connection for an external AI tool, name it, and grant only selected Ready Tables. The URL and key are shown once. Change table grants at any time, rotate a key to stop the old one, temporarily disable a link, re-enable it, or delete it permanently. Tables links are separate from the MCP servers used for Data Room documents.

02

At a glance

ProblemWhat to checkNext step
Analysis skippedAI analysis was unavailable or the organization reached its token limit.Review and complete the basic proposal manually
Import refusedFormat, size, row, column, or sheet selection does not meet the import limits.Use CSV, TSV, or Excel under 25 MB, 50,000 rows, and 200 columns
Agent cannot find a TableThe Table is not Ready, visible, or granted to that connection.Check all three access conditions
Agent selects the wrong TableThe description or agent instructions are too broad.Clarify what the Table contains and when to use it
Results look outdatedThe re-import may not have finished or become active.Confirm the newest version is Ready and active

Important

Document Markdown encryption does not cover Table rows in this release. When document encryption is enabled, Command displays this limitation during import and requires acknowledgement before publishing.

03

Related