Claude on agents

Hi guys,

Maybe a bit of a stupid question but i’m new to using external LLM’s inside of cursor.
I really love the cursor AI window, the possibility to select the modes etc. As most of y’all know opus drains tokens and i don’t have the wallet to pay 50 bucks a day for opus use. Especially when claude code max is a 100 a month.

My question is, instead of opening it with an extension, can i use claude code inside the cursor ai window? As a model. Just like you can overwrite a key for model usage.

If you got any suggestions on different options please let me know. Want to stick with cursor if possible ofcourse.

~ Billyisaliv3

I would love this ability as well. I had to switch to Claude Code Max to get essentially unlimited Opus use and, while I use it inside Cursor (both CLI for new --chrome extension plugin and the VSCode extension when I can) it obviously doesn’t feel as nice as an integrated solution with inline-diff and such.


This is a pretty serious threat to Cursor. I dropped from $200/month sub to $20 as I now only use Auto-complete and Grok Code for git-commit slash-commands and getting in to debug mode (which I then pass off to Opus in Claude-Code to fix things).

With all of the UX issues in Cursor, I am starting to consider other options. Auto-complete is really the only thing keeping me around;

  • The browser has been broken in several ways since before 2.x released and even if you use it in the tab in the main window, it is a poor experience (even if it could scroll). The Claude chrome extension and plugin is amazing; way better than chrome dev tools MCP. I’ve been posting here since the browser showed up asking for a way to connect to any chrome instance…well now Claude can do it with Opus flawlessly. I also don’t see much reason for a designer-flow when I can just ask Opus to make changes faster and faster

  • Inline diff is nice, but the built in VSCode git diff works almost as well. it is also live in the editor and I can approve blocks (stage them and they disappear from the diff), revert things, still view accepted diffs by clicking on the staged file name (actually a feature Cursor doesn’t have really, though this git flow can be used with it, so no big deal). Again not quite as good as inline diffs, but I’ve gotten used to it and don’t mind it.

  • I don’t really care about switching models since Opus is the best and even running parallel subagents doing tasks in 10 work tree branches doesn’t max out my usage at $100/month. I could use Sonnet for some things, but why bother? I use GPT-5.2 for some debugging help if Opus gets stuck, maybe 3% of debug sessions.

I am sharing this because I love Cursor and I know they are under pressure to fix UX and stop releasing broken versions (they acquired a code-quality team!). The Agent chat IS the best (when it works), but Claude Code is just so so solid and can do so much! …I can even click a file path to review it and open the file, not a review-view (sorry, couldn’t resist needling on this). …and I just realized the only reason I use Cursor for the git commit flow is an accident. If I do not add git commit to the commands allow list, it is forced to leave me in an editable shell. If I could get Claude to do this I wouldn’t use agent chat in Cursor at all right now (this can change of course…fast…GPT-6 maybe one day?)

tldr; Right now I am only using Cursor for editor tab IDE stuff and my git commit slash-command and spinning up and cleaning up the occasional debug session. Opus in Claude Code, running inside Cursor 95% of the time, is my constant assistant, all day long, with no worry of running out of usage.

The only reason I tried Claude Code last month for the first time was because Opus burned through my $200 cursor sub 3 days early and I had a deadline to hit, and Cursor support wouldn’t let me pay another $200 to start a month early. They didn’t take my money, so I accidentally found a way to do far more at half the cost - fixed cost.

I’ll keep posting feedback and trying new features. I look forward to all of the current issues being fixed. I’d be happy to be won back over fully.


If I could use my Claude Code subscription in Cursor Agent chat, well that would be pretty awesome.

I 100% agree, after the 2nd month of struggling i switched over to antigravity just to check the usage each 5 hours. Which should already give me way more play compared to the Opus usage on cursor.

Hopefully they soon fix their pricing, collaborations or anything that helps with this situation. Actually would love to stay on cursor.

The pricing for Cursor is currently very unclear; you can easily run out of tokens if you’re not careful.

Claude Code also does not anymore allow usage of their subscriptions in other products.