January 30, 2026

Despite the Claude Code hype, I still use Cursor

I've been a DAU of Cursor since last March and continue to be. Claude Code is very hyped. After trying it out, I think the appeal of Claude Code comes down to these four factors:

  1. Opus 4.5 is an insanely good model for coding and knowledge work generally.
  2. Claude's pricing is relatively generous. Precise token limits by plan are hard to determine, but if you use hundreds of millions of Opus 4.5 tokens per month, the Claude Max 20x plan is the cheapest way to consume them.
  3. The terminal UI makes you feel like a 1337 h4x0r, even though it's essentially a chat interface. This is an underrated aspect of the appeal and buzz in my opinion (see this tweet).
  4. Claude skills and plugins are apparently useful though I admit I haven't perused those libraries yet.

I've been using Cursor as my daily driver for almost a year now, and I don't plan to switch to Claude Code, at least not yet. Here's my thinking, starting by addressing the above reasons one by one:

  1. Opus 4.5: You can use it in Cursor.
  2. Price: This is the best reason to switch from Cursor to Claude Code. I'm at about 700M Opus 4.5 tokens consumed in my current billing period. Google AI Mode estimates I'd save $200-400 per month by switching to Claude Code. That's a real point in Claude's favor.
  3. Terminal UI: Doesn't matter to me. If it did, Cursor CLI exists now too.
  4. Claude Skills: You can now use them in Cursor.

So price is really the main reason I'd switch. But Cursor has a number of advantages to me over Claude Code.

  1. IDE interface: I tend to make changes to front-end files because I have opinions (right or wrong) about details of the UI. I also like being able to review files in a familiar IDE environment to see what's going on or look at diffs.
  2. Model-agnosticism: Cursor gives you access to Anthropic's models (Opus 4.5 is the main one I use these days), but also everyone else's. I use Composer-1 for straightforward implementation tasks because it's usually faster. I'm interested in trying out Codex-5.2 because apparently it's the best at debugging. New powerful models come out from different companies every few months. I like the model optionality that Cursor provides.
  3. Bugbot: I'm a huge fan of bugbot, which has saved me on numerous occasions. It now has an "autofix" mode, in which bugbot commits fixes directly to a branch after identifying issues. So good.
  4. Switching cost: I've already configured Cursor. I've set up integrations, MCP connections, rules, extensions... Yes, I can set all this stuff up in Claude Code, and yes it's pure laziness to not want to do so, but I still don't want to do it.
  5. Cloud agents: These exist in Claude Code too, I think, but I like being able to kick off a Cursor cloud agent from Linear or from my phone and get a good PR.

I'm effectively paying a few hundred dollars every month for these benefits of Cursor over Claude Code. But to me it's worth it.