Back to the changelog

• Security

  n8n executions: only for flow designers, without confidential data

  The detail of an n8n execution linked to a case is now visible only to flow designers or higher roles. It shows only what is needed to understand what happened, with credentials and the values of the case's confidential fields hidden. Content too large to check is replaced by "[truncated]". If an integration read these executions with an Operator user, it will need a higher role.

• Security

  Confidential data is no longer overwritten by placeholders

  When an Event or an automatic task response tries to write into a confidential field the text FlowSharp uses in place of hidden data, that value is ignored and noted in the history, instead of replacing the real data.

• Improvement

  Participant contacts more useful to the AI assistant

  Language, preferred channel and external reference of participant contacts are not personal data and now reach the AI assistant in clear, which reasons better about cases. Values that really identify a person stay protected. In conversations already started, external references protected earlier stay hidden until a new conversation.

• Improvement

  Plainer MCP errors when data is turned off

  When an MCP key cannot read case data, errors no longer suggest widening sharing: fixed messages in Italian and English with a clear code. A hidden flow and a missing flow get the same answer. Technical notices, such as the start confirmation request, stay visible.

• Improvement

  Read log stable under load

  The log of case and document reads now handles many simultaneous reads by the same person without errors and without slowing the rest of the system. If it cannot record, the read is still refused, as before.