hey everyone!
we’ve just made some changes to our documentation at https://docs.cursor.com and we’d love to get your feedback on it
want to make sure we cover as much as possible and is useful!
Please let us know where you feel the biggest gaps in knowledge are on Cursor so we’re sure they get good coverage in the docs!
Would be great to have some examples around the .mdc files.
Marketplace/Extension documentation is missing.
See here for an explanation and documentation example:
Personally, one thing I never understood (and also can’t find info about this in the documentation) is how context handling works exactly.
Let’s say I’m using the chat and providing a few files as context along my message. Do I still need to reference these files in my next message if I want to continue working with them, or do they remain relevant throughout the entire conversation? My experience regarding this has been inconsistent - sometimes I don’t need to provide the file in order for the llm to still remember it (makes sense, I assume its relevant parts are still in its context), but sometimes it can’t see them even if I make sure to attach them.
I’m also unsure how much I need to worry about overloading the llm’s context, as I don’t know what happens to these attached files under the hood. For this reason I found myself using the codebase search (ctrl+enter in chat) less and less as I don’t want to confuse the llm and drown out the previous messages. I would like to know if all that happens is that all of those found files get thrown into the conversation, or if there is some more work going on behind the scenes to only include the relevant parts.
It’s also inconsistent when it comes to handling any changes I have manually made between messages - if I apply its suggestion, change a line of code, and then ask it to tweak the code further, its next suggestion will omit the change I’ve manually made, even if I attached the file to my new message, leading me to believe that the llm only sees my old code that I have attached in my first message. Does that mean that I shouldn’t even bother including a file in every single message of mine if I have already included it once?
Generally, I would appreciate a little more transparency when it comes to how this context handling works under the hood, not just with files but with other @-symbols. I understand if you can’t reveal everything, but right now it’s too big of a mystery for me which can be infuriating at times.
Google Gemini is missing from the models page 
This isn’t really feedback about gaps in knowledge, but I’ve noticed, just by skimming through the pages, that there are a few typos. It’s worth doing some proofreading.
For example, on the very first page in the docs (Cursor – Welcome to Cursor), there is a typo:
Cursor provides the user with a few fundimental features that are only made possible by the development of LLMs (Large Language Models).
The bolded text should read fundamental.
Also on the page about Cursor rules (Rules for AI):
You can create a new rule using the command palette with
Cmd + Shift + P>New Cursor Rule. By using project rules you also get the benfit of version control since it’s just a file
The bolded text should read benefit.
This is just 2 typos I found in the span of 1 minute of skimming through the docs. Maybe I got lucky and found the only 2 typos in the entire documentation, but it’s still worth re-reading and fixing writing errors like those.
Should include Agent function calls like read_file and their params and any other relevant stuff (like the 250 lines limit) I think any Dev using the tool should know what its limitations and functionalities are
Linux installation should be implemented similar to windsrf please. Make it easy.
Thanks for catching these! Fixes on the way!
Ran each page of docs through LLM to catch issues - collated them here: https://docs.google.com/spreadsheets/d/1eHM7MzXI3aDNzWDDxP0PgdmrIV5E8h9_YPRkRwk2Vtc/edit?usp=sharing
I’ve shared access with @ericzakariasson
The only examples of MDC files I found were here.
Used it to make this gist, but there really should be clarity on what goes INSIDE .mdc files.
Japanese users are looking forward to Japanese documentation. Now that the website is available in Japanese, we hope to see the official documentation in Japanese as well.
Agreed. I think context management should be much more of a first class citizen of the app. The UI for adding and removing files to the context kinda … leaves a lot to be desired. I think context bundles would be nice. As would an easy button to add all currently open files to the context.
You have the following forum post:
This information really needs to be included here: Cursor – Usage
This does not convey the information:
users who run out of fast premium credits will be moved to a slow pool, which essentially is a queue of users waiting for a fast premium request to become available.
They are saying two different things.
An introduction to setting up a WSL environment (with the necessary extensions) would be useful, and an equivalent page on Microsoft Learn might suffice. WSL で VS Code の使用を開始する | Microsoft Learn
It’d be great to clarify how Cursor handles files that contain secrets. Link to convo: CRITICAL: Cursor does NOT follow .gitignore and .cursorignore files! - #6 by kaandok
I think we are in a spot where everyone needs a lesson on secrets handling.
for example, im yolo-ing the heck out of sidelings, which always pull from a .env but my yolo has full GH access… and while im doing things in complete sandbox… im still at total risk.
a complete howto on secure secrets within use of Cursor is needed.
We talk . ruels and such, but we really need MasterClass on secrets within @cursor…
Refund related queries could be communicated in docs and also wherever they apply for refunds