DATA MASKING / WORK ON A COPY

Change the values.
Keep the source.

Create a separate dataset for testing or demonstrations. Choose what to mask, inspect the result and save the reviewed copy as JSON. Source rows stay unchanged.

Explore masking methods Download trial ↗

New in the development build · Public download coming soon

Data masking rules and preview with substituted names, hashed emails, shuffled phones, varied amounts and NULL notes — Development build · Fictional data in disposable SQLite · 22 October 2026

FIVE WAYS TO CHANGE A VALUE

A rule for each column.

Open Ctrl+K → Data masking, choose your tables and set a method for each eligible column.

Substitute.

Replace text with a pattern. Use {n} for a stable number per distinct value.

Vary.

Adjust numbers within a chosen amount, or dates within a chosen number of days.

Shuffle.

Rearrange existing values within the selected column.

Hash.

Generate keyed hashes with a fresh secret for each generation. The secret is not exported.

Clear.

Set eligible nullable fields to NULL when you do not need their values in the copy.

PREVIEW BEFORE YOU SAVE

Inspect the copy.
Then take it with you.

Generate a masked preview, review every included table and confirm the fields and values before saving. The JSON export uses that exact generated dataset.

Leave the source in place.

Masking reads source rows and builds a separate dataset. It does not update the original tables.

Keep keys intact.

Primary keys, generated columns and declared relationship columns cannot be masked. There is no key remapping.

Review the whole copy.

Unselected fields remain unchanged. Hashing and shuffling alone do not guarantee anonymity; check the data you intend to share.

What is included in the export?

The reviewed data and structure metadata in JSON. It is a data copy, not a complete schema backup. Database collations, checks, defaults and triggers are not recreated or fully validated.

Which databases and dataset sizes are supported?

PostgreSQL, MySQL, SQL Server and SQLite. MySQL requires InnoDB for snapshot guarantees. Select up to 12 tables and 50,000 rows total, with 256 columns per table, 12 MiB input and 16 MiB compact output. The preview shows the first 25 generated rows.

Does masking need an AI provider?

No. These masking methods do not use AI. Fill with AI is a separate workflow for generating reviewed column values.

Prepare a copy
you can inspect.

45 days to explore. £5/month after your trial. Public downloads are being prepared.

Download trial

Change the values.Keep the source.