I use Claude with Jira and Confluence most workdays. Atlassian's Remote MCP Server respects the permissions already attached to my account. I ask for the open tickets on a program, a first draft of a sprint summary, or Jira issues pulled from a long decision thread. Those are ordinary tasks, which is why the integration has lasted longer than the demo phase.
Jira and Confluence hold the history I would otherwise have to reconstruct: what was tried, what was blocked, and which decision changed the plan. Claude often finds a ticket or comment I missed. It gives me a better starting point for review, but I still need to read the source and decide what matters.
What I have stopped doing is more useful than a feature list. I no longer write status updates from a blank page, hand-summarize every sprint retro, or chase stale tickets one by one. I use the time to push back on scope, find the assumption nobody recorded, and make the call the ticket history cannot make.