Cleaning a table

Clean data while preserving the source

Open an ordinary Table and choose Clean data from its actions. Loam builds a closed, deterministic plan and runs a bounded preview before anything is saved. The original Table is never edited or replaced.

Supported steps

• Choose the projected columns.
• Rename a selected output column.
• Filter with scalar comparisons or null checks.
• Remove rows that are exact duplicates.
• Normalize text to lower or upper case.
• Convert a timestamp column to a date.

Loam does not accept arbitrary expressions in a cleaning plan. Fuzzy matching and whitespace rewriting are available only in the separate browser-local CSV utility and are not silently substituted here.

Preview and save permissions

A Viewer can profile and preview an accessible Table. An Editor can also save the reviewed plan as a new non-materialized View beside the source. A root-level or system-managed Table does not offer a save destination.

Use the same tools from Claude

The cleaning tools are part of Loam's existing MCP connection. Add the ordinary Loam connector to Claude, then ask it to profile a Table, find exact duplicate groups, preview a plan, or save a reviewed cleaned View. Saving requires separately consented write access plus current Editor authority.

No minisite backend

The public CSV cleaner runs in its visitor's browser. It has no regional database, engine, MCP endpoint, or data proxy. Signed-in data cleaning happens here in Loam through the same regional APIs and permission checks as other Table work.