Screens: AI, MCP and security

Column masking

Column masking shows sensitive values masked to team members of the roles you choose in an account project. Set the roles to mask in the Project Configuration window. Specify the values to mask in two ways:

  • Designate a sensitive column to mask its entire value.
  • Register a regular expression pattern to mask only matching text.

Masking controls only the values you see; it does not encrypt data. That limit is explained in Security and where your data goes.

Where you can use itWebDesktopWithout signing inSigned inLocal projectsAccount projects

What's on the screen

Masking is set per role in the Project settingsRole Permissions tab. The Project settingsRole Permissions tab is visible only to the owner of an account project and to roles with All Permissions turned on, and only the owner changes its values.

No.NameDescription
1Role tabsOpen the Manager, DBA and Member roles one at a time. The owner is not in the list and always sees the original values.
2Sensitive Data Masking checkboxWhen checked, team members of this role see sensitive columns and pattern matches masked. It cannot be checked for a role with All Permissions turned on.
3Column MaskingLights up with a count when sensitive columns are designated. Click the count to see the list of connections and columns.
4Pattern MaskingLights up with a count and the pattern names when patterns are turned on.
5Manage Patterns… buttonOpens the window for registering and editing patterns (see 'Registering a pattern' below).
6WarningText describing the limits of masking. It is the same as the box below.
  • It is turned on per role, but it carries down. Masking Manager also masks DBA and Member, and masking DBA also masks Member. A checkbox turned on because of a higher role cannot be changed and is marked (automatically applied because a higher role is masked).
  • A masked role gets both sensitive columns and patterns.

Choosing the roles to mask

  1. Open the account project with the owner's account and open the Project settingsRole Permissions tab.
  2. Open the tab of the role to mask and check Sensitive Data Masking.
  3. Save the Project Configuration window. It applies to SQL tabs that team members of that role have open from their next run.

Designating sensitive columns

Columns can be designated only when at least one role is masked. Right-click a column and choose the Mark as sensitive menu.

  1. Right-click a column in one of the following places.

    WhereColumns you can designate
    A table node in the ERDCharacter-type columns of a table that exists in the database, excluding primary and foreign keys
    The column grid of a table tabSame as above
    The column grid of a view tabEvery column of the view, regardless of type or key
  2. Choose the Mark as sensitive menu (1). The designation is saved to the server immediately; there is no save button to click.

  3. Designated columns get an orange lock icon before the column name in table tabs and view tabs.

  • Example: in the demo shop, the email and name columns of the customer table and the line1 column of the address table are good candidates. customer_id is a primary key and address.customer_id is a foreign key, so the menu does not appear for them.
  • The owner and team members of roles that are not masked can designate columns. Only the owner can undo it (the Unmark sensitive menu), because undoing makes the values visible again to masked roles.
  • Tables and columns newly created in the ERD and not yet applied to the database cannot be designated.
  • If a designated column later becomes a primary or foreign key, the designation is removed automatically when the owner opens that table, because rows cannot be edited or referenced through a masked key.

Registering a pattern

Patterns form one list for the whole project. They mask only the parts of string values in results that match the regular expression, so use them for phone numbers or email addresses mixed into sentences such as notes and descriptions.

  1. In the Project settingsRole Permissions tab, open a masked role and click the Manage Patterns… button. The Manage Masking Patterns window opens. The switches in the list turn patterns on and off.

  2. Click the + Add Pattern button.

  3. The buttons in the Quick start: row (on the English screen Mobile Phone (KR), Mobile Phone (US), Email and Credit Card No.; the list differs by UI language) fill in the name, regular expression and sample text. To write your own, fill in Pattern Name and Regular Expression.

  4. Choose Masking Mode.

    ModeHow it masks
    PartialKeeps a few characters at the start and end of the match and replaces the rest with the character in Mask char. With Keep separators (- space) checked, - and spaces are kept
    FullMasks the whole match
    TemplateReplaces it with the text in Replacement Template. Use $1, $2 to insert the regular expression's groups
  5. Enter sample text in Preview to check the result, then click the OK button.

  • Values filled in by the buttons in the Quick start: row (1) can be edited. If the regular expression (2) is invalid, the preview (4) shows Fix the regex to see the preview..
  • The option fields below change with the mode (3).
  • Only the owner and roles with All Permissions turned on can register and edit patterns. Patterns are saved directly in this window, separately from saving the Project Configuration window.

Where masking applies

Values are masked by the server that runs the query, before the results are sent to the screen: on the local server on the team member's device for the desktop app, and on the NeoSQL server for the web. The following is what happens for a team member of a masked role.

PathMasking
Result grid of the SQL editorSensitive columns show **** and pattern matches show the configured form. Copies stay masked, and masked cells cannot be edited (Result grid)
Exporting results, the Data Export... item in the explorerMasked values are written to the file (Exporting results)
Native BackupDump tools cannot mask values, so the menu is hidden and the server (the local server on desktop, the NeoSQL server on the web) rejects the request (Data export, import and native backup)
Rows AI Assistant queriesMasked by role. Turning on Always mask sensitive data for the connection masks regardless of role
Results of execute-query from an external MCP clientSame as AI Assistant. The policy is read again from the server for each run, and the query is not run if it cannot be read
RLS simulatorNot masked. See 'Things to keep in mind' below

Things to keep in mind

  • Masking is not a database permission. Values can come out through detours such as views or functions that return the same value, or queries that filter on the value (WHERE email LIKE 'a%'). For values that must stay hidden, encrypt the column or split database account permissions on the database side.
  • The RLS simulator opened on PostgreSQL connections does not mask results. Team members of masked roles can see original values of sensitive columns through the simulator (RLS matrix and simulator).
  • In the desktop app, both masking and the database connection happen on the team member's own device. Do not use it to protect values from someone who has full control over their device.
  • When roles or sensitive columns change, open SQL tabs of team members apply the change from their next run. A tab that missed the change notification may keep running with the old policy until the project is opened again.
  • As the If structure information cannot be read, some additional values may be masked or editing may be restricted. text below the warning says, when NeoSQL cannot work out which column of which table a query reads, it masks more or blocks editing to be safe.