Cursor: An Honest Review from Daily Use
I’ve been using Cursor for about six months now, both for personal projects and professional work. It’s an AI-powered code editor built on top of VS Code, meaning it inherits all of VS Code’s extensions, themes, and keybindings. But the AI integration is what sets it apart—and it’s both a blessing and a curse. Here’s my detailed take.
What It Does Well
1. Context-Aware Code Completion
Cursor’s inline suggestions are smarter than GitHub Copilot in one key way: it understands your entire project structure, not just the current file. For example, when I was refactoring a React component that imported a utility function from ../../utils/helpers, Cursor suggested the correct import paths and even offered to update other files that referenced the old function. It also handles multi-line completions well—if I start typing a function signature, it often guesses the entire body, including error handling, based on patterns in my codebase.
2. Chat That Actually Reads Your Code
The built-in chat panel (Ctrl+Shift+L) lets you ask questions about your code without copy-pasting. For instance, I asked: “Why is this API call failing silently?” It scanned the relevant files and pointed out that my try/catch block was swallowing errors because I forgot to rethrow. It also lets you highlight specific code blocks and ask for explanations, refactors, or bug fixes. This is far more useful than generic ChatGPT conversations.
3. “Apply” Edits Directly
When you ask for a change in chat, Cursor can apply the edit to your file with a single click. No copying, no pasting, no guessing. For example, I asked it to “add TypeScript types to this function,” and it inserted the types inline, highlighted the changes, and let me accept or reject them. This feels like a real workflow improvement, not a gimmick.
4. Agent Mode (Beta)
The “Agent” feature lets you give a high-level instruction like “Add a dark mode toggle to this React app, using localStorage to persist the preference.” Cursor will create new files, modify existing ones, install dependencies, and even run terminal commands. It’s not perfect—it sometimes creates unnecessary files or misconfigures imports—but for boilerplate tasks, it saves 15–20 minutes per task.
Limitations
1. Hallucinations and Overconfidence
Cursor will confidently generate code that looks correct but is subtly wrong. I asked it to “add pagination to this list component,” and it created a Pagination component that used a library I wasn’t using, with incorrect prop names. You must review everything it writes. It’s a tool for speeding up, not replacing, your thinking.
2. Performance Overhead
On a mid-range laptop (16GB RAM, i7), Cursor can slow down noticeably when working on large projects (100+ files). The AI suggestions lag, and the editor sometimes freezes for 1–2 seconds when you type. It’s not unusable, but it’s not as snappy as plain VS Code.
3. Pricing Reality
The free tier gives you 50 “slow” requests per month (they take 5–10 seconds) and 500 fast requests. The Pro plan is $20/month, which gives unlimited fast requests and access to GPT-4 and Claude. The Business plan is $40/user/month for team features. For a solo developer, Pro is worth it if you use it daily. But the free tier is too restrictive for real work—you’ll hit the limit in a few days.
4. No Real “AI-First” Features
Despite the hype, Cursor is still a text editor with AI bolted on. It doesn’t automatically refactor your codebase or suggest architectural improvements. You have to explicitly ask for everything. It’s not “autopilot”; it’s a better autocomplete.
Key Workflows
Debugging: When I’m stuck on a bug, I highlight the suspicious code block, open chat, and ask “What’s wrong here?” It usually spots missing edge cases or logic errors.
Refactoring: I use the “Edit” command to rename functions or extract components across multiple files. It handles renaming props and updating imports reliably.
Learning: I often paste unfamiliar code (e.g., a complex regex or a new library snippet) and ask for a plain-English explanation. It’s excellent for this.
Who Should Use It
- Solo developers who want to speed up repetitive coding tasks (writing boilerplate, adding tests, fixing lint errors).
- Intermediate coders who can review AI output critically. Beginners might trust it too much and ship broken code.
- Teams who already use VS Code—Cursor’s compatibility means zero migration cost.
Avoid if: You’re on a tight budget (free tier is too limited), you work on highly sensitive code (Cursor sends code to its servers), or you prefer a minimal editor without AI distractions.
Verdict
Cursor is the best AI code editor I’ve used, but it’s not a revolution. It’s a polished autocomplete with a chat window. For $20/month, it saves me a couple of hours per week. That’s worth it—but don’t expect it to write your app for you.
