Is there any way to restrict the models that are allowed to be used by everyone in a business plan? For instance, Only allow Claude-Sonnet-3.5 and Claude-Haiku-3.5
Not currently unfortunately, but we hope to improve the configuration around this in the future!
Can I ask, is there a reason you want to limit the models your users can use? Knowing the use case may help us build something around this that works best!
Hi Dan, thanks for the reply!
We have internal guidelines set by our security team that has an allow list of approved models to work with. Some of the models provided by Cursor are not approved, so it would be great to be able to be able to set which models are used to remain compliant.
Makes sense, will feed this back to the team!
Want to second this! Legal/security at my org also aren’t comfortable with all of the models and we’d like the ability to use only those which we’ve reviewed.
This is absolutely needed now after introducing gpt-4.5 where in an enterprise situation the team is using it unknowking of the costs and blocking the usage limit for everyone else incvluding themselves for other models!!
We need this as well! Internal security/compliance requirements prevent us from using deepseek and in order to use Cursor we need to restrict that model.
This is a pretty obvious control. Companies will not want each developer to have the option to completely subvert their data use and protection requirements and then trust that only vetted and authorized models are actually in use. Different models have different use licenses, data protection standards, and obviously processes data in different locations. Some or all of these factors can limit which models can be used. If we can’t control where our data goes then we can’t use the platform that enables those unwanted data flows.
What’s the feedback and timing from the team? This will factor in to our decision to continue using.
@danperks
Is there an update about this one?
Is this feature included in the roadmap?
No update yet, it’s something we’ve tested internally but is needs some significant work before we can make it a feature for the business plan outright.
Are there any workarounds or best practices for ensuring users adopt specific models—aside from simply telling them to? If not, no worries; I’m just checking whether you’ve seen any alternative approaches.
I’d recommend contacting our sales team via this form, as they can help confirm what options are available to you regarding this: Contact our sales team | Cursor - The AI Code Editor
+1 on this feature request.