Table of Contents
Quick Answer
Cursor slows down when it indexes large codebases, holds too many files open, or your LLM calls are congested. Fix by adjusting indexing settings, limiting open tabs, and using the right model for the task.
- Exclude heavy folders from indexing (node_modules, .next, dist)
- Close unused tabs and editor groups
- Use smaller models (Sonnet, Haiku) for simple tasks
Why This Happens
Cursor is a VS Code fork with deep AI integration. It indexes your codebase for context-aware completions and chat. Performance degrades with: massive repos (>500MB), too many files open, outdated version, heavy extensions from VS Code migrated over, or slow network to AI APIs.
Step-by-Step Fixes
Step 1: Update Cursor
Cursor → Check for Updates. Monthly releases include performance fixes.
Step 2: Exclude heavy folders from indexing
.cursorignore in repo root. Example:
node_modules/
.next/
dist/
build/
.turbo/
coverage/
*.log
Step 3: Close unused tabs
Tabs keep file content in memory. Right-click tab → Close Others.
Step 4: Close extra editor groups
Too many splits slow rendering. Use one editor group.
Step 5: Disable heavy extensions
Extensions → Disable the ones you don't use daily. ESLint, Prettier, GitLens can be memory-heavy.
Step 6: Use the right model
- Quick autocomplete: Cursor Fast (cheap, fast)
- Simple edits: Sonnet
- Complex tasks: o1, Claude Opus — but slower
- Don't use Opus for "rename variable"
Step 7: Clear chat history in composer
Long composer threads slow down. Start a new one.
Step 8: Check network latency
Cursor → Settings → Network. High ping to Cursor's API slows suggestions. Wired connection helps.
Step 9: Reduce open file context
In Settings, lower "Max context size" if you don't need it. More context = slower responses.
Step 10: Restart with clean state
Close Cursor completely. Reopen. If still slow, disable all extensions and re-enable one by one.
Step 11: Check system resources
macOS: Activity Monitor. Windows: Task Manager. If Cursor uses >90% CPU for long periods, file a bug.
Step 12: Use .cursorrules for scope
Project-specific .cursorrules tells the AI to focus on relevant areas, reducing context load.
When to Contact Support
- 100% CPU persists after clean restart with no extensions
- Indexing never completes on medium-size repo
- Crashes on startup after update
Support: cursor.com/support or forum.cursor.com.
Prevention Tips
- Maintain .cursorignore from day one of each project
- Restart Cursor weekly to clear memory
- Monitor model costs — heavier models slow + expensive
- Keep your OS updated (macOS, Windows security patches matter)
FAQs
Why does Cursor use so much RAM? Large codebase indexing + open files. Exclude heavy folders.
Is Cursor faster than VS Code + Copilot? For AI features, yes. For pure editing, roughly the same.
Does .cursorignore affect chat context? Yes — ignored files are excluded from indexing and chat.
Can I run Cursor offline? Editor works; AI features need internet.
Why does Composer freeze? Very long threads. Start a new composer.
Does Cursor work with monorepos? Yes, but index selectively per app.
What's the best model for code review in Cursor? Claude Sonnet 4.5 or GPT-4o balance quality and speed.
Conclusion
Cursor speed depends on configuration. With exclusions and the right model per task, it's fast. For AI-powered coding across multiple models without vendor lock-in, try Assisters AI.
[Try Assisters AI Free →](https://assisters.dev)