Claude code vs Cursor?

Tried using Claude code and i found it has improved a lot and cheaper too. whats your take guys?

I am also curious to hear it. I’m currently using Cursor IDE, and then I heard a lot about Claude Code in the terminal and how people praise it. Which led me to discover about Cursor CLI. It makes me wonder, what is the difference between the cli and the ide?

And what is the benefit of using the cli and how do people compare it to Claude code? I couldn’t test Claude code, but I’m also not a fan of changing tools so regularly.

The the recent spruce code leak we could get a glimpse of how Claude code works under the hood (especially thanks to https://ccunpacked.dev ), which then led me to wonder, what is the loop of cursor cli?

Claude Code is, by design, the most cost-effective way to use Anthropic’s models. That’s not a coincidence - Anthropic built it that way. No other IDE or integration will give you the same value per dollar when working with Anthropic models.

Beyond pricing, there are meaningful functional differences that are exclusive to Claude Code. The tool harness Anthropic built for it is genuinely superior - not dramatically so, but noticeably and objectively better in the way it handles context and orchestration. It supports multi-agent workflows in a way that Cursor simply doesn’t replicate. The only models worth reaching for right now (4.5+) are best accessed through Claude Code, and you get capabilities there that just aren’t available elsewhere.

Cursor used to offer the best IDE interface in the business, and that’s exactly why people like me were paying stupid amounts of money for it. The UI was the product. But the latest release has completely abandoned what made it great. The new interface is a significant step backward (or maybe a leap sideways is a better analogy) - cluttered, unintuitive, and seemingly designed to chase a different audience & compete with OpenAI (laughable decision). I cancelled my Cursor subscription before the update, and seeing this confirms it was the right call. At this point, VS Code offers a better version of the experience long-time users like me enjoyed with Cursor 1 and 2.

To anyone at Cursor reading this: your users chose you because of the interface you built, not in spite of it. Pivoting to imitate OpenAI’s Codex-style tooling - which is aimed at non-developers - is a strange direction for a product that had genuine best-in-class status. I hope it’s reconsidered, but I doubt it will be.

For now, I’m putting my money with Anthropic. They’re the only ones who actually seem to understand what real users need - users who aren’t just vibe coding 40k lines of slop per day.

3 Likes