Cursor IDE
Hello,
In my company we are using cursor TEAMS.
I am the one that writes the cursor rules , skills, mcps…
The rules are key and we have a lot of Intellectual Properties we would like the users to not even be able to read. The idea is to protect those rules as I do not want them to go to the competition.
How can we achieve that? Do you plan to install that? It is a big deal for us…
Thanks
Hey @jbesclapez, thanks for the suggestion, and for explaining the use case.
Today there is no way to hide or encrypt Team Rules from members.
Rules are instructions the agent reads on each member’s behalf, so they are visible in Settings and in the dashboard, and anything the model receives can also be surfaced by the person using it (for example by asking the agent to repeat its instructions). That applies to any AI coding tool, not just Cursor, so true “use but cannot read” protection for prompt text is not something we can guarantee.
What teams usually do for genuinely proprietary logic is keep it out of the prompt entirely: put it behind an MCP server or internal API that you host, so the agent calls a tool and only sees inputs and outputs (keep in mind that tool outputs can be visible to users), while the actual know-how stays on your infrastructure. Team Rules then only need to tell the agent when to use those tools.
Hi @Colin and thanks for having a look at it.
As you can imagine, I am not satisfied with that solution - your answer is correct, dont get me wrong. But we are externalizing the solution to an MCP.
Ideally I would like to have a password protected rules/encrypted and an api that checks the password validation of some sort… I am not into security of those things you can imagine! 
I hope Cursor Team will make this happen at some point…
Thanks @Colin