Cursor plans must be in .cursor/plans/ to appear in the Plans panel and be executable through the UI. If you move them elsewhere, they become inert markdown files — Cursor won’t recognize or run them.
We like to check in plans and other files as a sort of “source code”, and we would want to be able to run those plans from other folders besides the current user’s .cursor/plans/
What’s the goal behind checking plans into the repo in a different location? Is it about collaboration and sharing plans across team members, version-controlling them alongside your source code, or something else?
Better collaboration and sharing of plans is on our roadmap, and it would be great to hear a bit more about what you’re trying to do.
For me it’s just organization. The plans are somewhat self documenting, keeping them in the project makes it clearer what they refer to. Having an ISO-8601-like (maybe just minute level accuracy) filename prefix (or just some control over naming) would also help keep them organized.
I generate plans for every agentic action that could write a file, as this greatly aligns outcomes with my expectations and keeps agents from wreaking havoc if I’ve mis-specified something. Some plans are essentially ephemeral, and others are for tasks that may require iterative participation from me over hours or days or longer (eg, interviewing myself or ingesting files for a document library). For longer-duration plans, it would make pausing and re-starting those tasks much easier if I could save/bookmark/store them in a way that was more organized and accessible than simply keeping the agent window or plan open in the application. If I accidentally close something, it can be a PITA to find it again.
I’ve built a significant knowledge base for my project, and agents regularly suggest extension to-dos for action items that are good ideas but I don’t want to execute at the moment. That’s the kind of thing I would want to store in a project/repo folder for ease of later access. I could theoretically have dozens of plans in progress for various ideas, and that quickly becomes a nightmare in the IDE.
As long as we’re talking about organization, we really need some better filtering on agent window history. Tags so we can group and filter by topic/target, maybe hierarchical mapping to keep track of sub-agents working off a plan, fork, or other predecessor. Pin and time sort only is not sufficient.
Edit: Somehow I had not yet discovered the “save to workspace” feature. This goes a long way. Oops. Pretty hidden in the UI, though.