Hi!
I have been using Cursor for a great while and I have to say that I really like using it! Just that I can use my own company’s own OpenAI proxy for LLM models in Cursor into its core UI through custom OpenAI API keys combined with the strength of having VS Code extensions for development in general is awesome on its own merits.
This made me think whether it is possible to achieve something similar when it comes to providing own vector databases. As I am writing this post, all the embeddings and so forth are stored on Cursor’s servers. While the documentation states that the chunks are encrypted, this still does raise skepticism among my colleagues that are concerned about safety and integrity of our codes being stored outside of the company. I was thinking that if Cursor could potentially allow me to configure a vector database into which all embeddings would be stored, that would have meant that parts of our codebase really never leave the company’s site and this would have changed the discussions regarding safety/integrity to Cursor’s favor at my workplace. Cursor could be that AI-first editor that glues all these parts together in a seamless elegant way. I am not sure if this is technically possible or if this is aligned anyhow with the expected development of Cursor for future, but at least I would like to get some feedback on this idea. Maybe it can be achieved today using MCP? I am all ears to hear all the tricks that you might have tried out and found to be successful.
Thank you all for your response in advance!