I built my first product, Voyage Bus Tickets (view it here), using just Cursor. Then I built my second product (which is being kept a secret) using both Claude Code and Cursor. Here is my opinion on whether Claude Code or Cursor is the better code editor.
Why I Tried Claude Code
After reading comments from several developers on X and YouTube saying Claude Code was the better coding agent and after not being able to figure out why in a quick manner, I decided to test Claude Code for myself.
Cursor Knew My Codebase Better Out of the Box
With Claude Code, I had to manually add instructions to CLAUDE.md explaining how types get published from the backend to the API client. Cursor understood this workflow immediately. Its codebase indexing had already e…
I built my first product, Voyage Bus Tickets (view it here), using just Cursor. Then I built my second product (which is being kept a secret) using both Claude Code and Cursor. Here is my opinion on whether Claude Code or Cursor is the better code editor.
Why I Tried Claude Code
After reading comments from several developers on X and YouTube saying Claude Code was the better coding agent and after not being able to figure out why in a quick manner, I decided to test Claude Code for myself.
Cursor Knew My Codebase Better Out of the Box
With Claude Code, I had to manually add instructions to CLAUDE.md explaining how types get published from the backend to the API client. Cursor understood this workflow immediately. Its codebase indexing had already embedded the repo.
How They Actually Work
- Cursor uses RAG-like indexing. It chunks your codebase, creates embeddings, and runs similarity search when you ask questions. Fast, automatic, no setup.
- Claude Code loads files into a 200K token context window. It doesn’t index codebases.
Cursor finds relevant code quickly. Claude Code maintains context across long conversations.
Where Claude Code Works Better
Claude Code offers far more context than Cursor per conversation. Claude Code loading entire files into context rather than using indexing seems to enable better performance on Claude Opus 4.5 (discussion about this on Reddit, more discussion on this from Mckay Wrigley)
The UX Difference
Cursor’s diff UI requires constant interaction. Every change brings "Keep All" or "Reject All" buttons. These appearing everywhere are a constant part of my workflow when using Cursor.
Claude Code edits files and continues, leading to better UX (in my opinion).
Extensibility
| Claude Code | Cursor |
|---|---|
| MCP servers (databases, browsers, APIs) | @-mentions (predefined) |
| Custom slash commands | .cursorrules file |
| Hooks | — |
| Skills/plugins | — |
The Subjective Part
I have more fun with Claude Code.
It’s not faster. Setup takes longer. It has no memory between sessions. But the terminal-native, conversational approach feels more like collaboration than code review.
When I Use What
| Cursor | Claude Code |
|---|---|
| Quick edits | Complex, multi-step features |
| Exploring unfamiliar code | Codebases where I understand everything |
| Zero setup | Custom workflows |
| Inline completions | Conversation over autocomplete |
Summary
I’ll keep using both. Cursor has great inline tab suggestions. For features where I want to maximize the performance of Opus 4.5 or work with an AI agent as an assistant rather than a bulk-edit tool, I’ll continue to use Claude Code.