Current situation
Currently, when using the MCP server to perform actions like updating a Jira issue description, the changes are applied immediately and destructively. There is no opportunity for the user to review the final payload or see a "diff" of what will be changed. In the event of a logic error (such as the tool overwriting the full payload and stripping attachments), the data loss is instantaneous and often irreversible through standard UI history.

Suggested improvement
We suggest implementing a "pre-flight" confirmation step for all AI-driven or MCP-driven write operations. This should include:

1. Visual Diff: A side-by-side comparison showing the current state versus the proposed state of the field(s) being updated.

1. Impact Warning: A clear notification if the proposed update will result in the removal of existing content, such as attachments, formatting, or inline images.

1. Manual Confirmation: A "Confirm/Apply" button that the user must click before the API call is executed, allowing them to catch errors before they go live.

Workaround:

We can set up a prompt/memory option in the connected AI tools to warn users before making any changes directly in Jira.