How reliable is Cursor for indexing documentation through its built-in tool?
I’ve read on Twitter, from several people, that it’s not that much, and that they prefer to scrap the docs elsewhere and just bring the Markdown to Cursor manually.
I’m wondering about the validity of these claims.
Thanks!
Hi @dny_ex I am using indexed Docs daily and its working without having to do anything with it.
What would have been the issue with ‘not that much’ and ‘prefer to scrap the docs elsewhere’?
I can see one clear issue with that alternative approach and thats: consuming more tokens feeding markdown from scraped docs unnecessarily and getting faster to rate limits.
Yeah, I totally agree with you. I haven’t “felt” that Cursor is not properly using the documentation that I’m feeding to it, but the difference can be so subtle that I couldn’t even tell if that were the case.
Like I see that Cursor indexed 450 pages of documentation “A”, for example, but what if it left out 20-30 pages? I could see myself totally saying that the LLM got stuck in a problem (or hallucinated) when in reality, it’s just missing the info from the docs that failed to index.
What would have been the issue with ‘not that much’ and ‘prefer to scrap the docs elsewhere’?
I saw someone saying that Cursor only indexed 1 page, so they preferred to go to Firecrawl and scrap, but again, that hasn’t been my case nor have I been checking regularly since I’m trusting that all documentation is getting properly indexed. Hence, the reason of coming here to see if any of you guys have experienced the same.
Yeah I dont necessarily have insight into docu indexing within Cursor. For stuff that I personally needed it worked.
Some are using separate MCPs for documentation.