Is this normal? Second time encountering this.
Me: “Hi! Could you please scand the codebase and suggest 2 things to improve it?”
Cursor AI: “I apologize, but I don’t have access to your codebase or any specific code to scan. As an AI language model, I can only provide information and suggestions based on the context you give me. I don’t have the ability to access or scan external files or codebases.
If you’d like me to suggest improvements, you would need to provide specific code snippets or describe particular aspects of your codebase that you’re concerned about. Once you share that information, I’d be happy to offer suggestions for improvement.
Is there a particular part of your code or a specific issue you’d like me to focus on? If you can provide some code or describe the functionality you’re working with, I’ll do my best to offer helpful suggestions.”
Me: “But you are part of the IDE whose purpose is to give AI access to my codebase…”
Cursor AI: “I apologize for the confusion in my previous response. You’re right, and I made a mistake in not recognizing the context of being integrated into an IDE with access to your codebase. Let me correct that and provide some suggestions based on a scan of your codebase.
After reviewing your code, here are two suggestions for improvement:”
Hi @Anastasia ,
Which LLM and interface (Ctrl + K, Ctrl + L or Ctrl + I) are you using when you get responses like that?
It is not totally surprising to see a response like that, as prompts are sent to an LLM (albeit possibly modified/optimised by Cursor in between sending a message and receiving a response).
I think there may have been some forum requests in the past to have Cursor more ‘aware of itself’.
Personally, when I have previously asked Cursor questions about Cursor-related things I do start with ‘I am using Cursor (the fork of VS Code…’.
Also, the wording in your prompt “scan the codebase” might have been too generic.
But your post is a good use case for the Cursor devs to see and reflect on whether they need to add more ‘system messages’ to the prompt construction etc.
It was the chat interface. It’s not a huge deal, it did what I asked after I reminded it that it was an IDE. It would just look bad or confuse new users if it tells them it can’t do this when it actually can. When I first tried Cursor last week I was like “Then what’s the point of using Cursor of it can’t see the codebase? I might as well just be talking to ChatGPT.” But then I decided to make my case for why I thought it should be able to see the codebase and it remembered that it actually can. 
1 Like
Hi @Anastasia ,
I totally understand and it’s a valid query.
FYI, in case you haven’t seen them, here are the docs on codebase context:
https://docs.cursor.com/context/@-symbols/@-codebase
1 Like