Tool Guide · Coding
Cursor AI Guide
The most popular AI-native IDE. A VS Code fork with deep model integration, multi-file edits, and a strong agent loop.
Key takeaways
- 1Multi-line tab autocomplete with codebase awareness
- 2Cmd-K inline edits with diff preview
- 3Cmd-L chat with @-mentions for files, folders, web, docs
- 4Agent mode for longer multi-file tasks
- 5Rules files (.cursor/rules/*.mdc) for project-specific guidance
- 6Indexed docs — @docs for any framework's documentation
- 7Privacy mode for sensitive code
- 8Restore checkpoints for agent runs that go sideways
Best for
- Day-to-day coding work in a GUI editor
- UI and front-end work where seeing the file matters
- Mixed autocomplete + chat + agent workflows
- Teams already on VS Code who want a familiar shell
Not for
- Server-side or SSH workflows — use Claude Code
- Very large codebases (>500K LoC) — indexing slows
- Workflows that need to span outside the editor
What it is
Cursor is a fork of VS Code where AI is a first-class part of the editor. You keep your extensions, keybindings, and workflow — but get inline edits, an agent mode, and tab-complete that actually understands your codebase.
Core features
- Tab autocomplete — multi-line, context-aware. Often predicts the next several edits, not just the next token.
- Cmd-K inline edit — highlight code, describe a change, accept the diff.
- Cmd-L chat — chat with the codebase.
@files,@folders,@web,@docsto scope context. - Agent mode — give it a task; it plans, edits across files, and runs commands.
- Rules —
.cursor/rules/*.mdcfiles inject project-specific guidance into every prompt.
Setting it up well
- Add a
.cursor/rules/general.mdcwith the rules of your codebase (testing, style, what NOT to do). - Choose your default model in settings — typically the strongest available for agent mode.
- Index docs: drop in URLs for your favorite framework's docs so
@docsworks. - Enable / disable Privacy Mode based on your data sensitivity.
When Cursor beats Claude Code
- You want a graphical diff before accepting changes.
- You're doing UI work and want the file open while you edit.
- You prefer mouse + keyboard over pure terminal.
- You want one tool for all coding (autocomplete, chat, agent).
When Claude Code beats Cursor
- The task is long-running and you want it in the background.
- You're SSH'd into a server.
- You want fine-grained hooks and MCP control.
- You want to script around the agent.
Tips
- Treat tab-complete as a multi-line proposal — don't accept blindly.
- Keep agent runs scoped. Massive multi-file refactors do better when broken into pieces.
- Use the "Restore checkpoint" feature when an agent run goes sideways.
@docsis criminally underused.
What it isn't great at
- Very large codebases (>500K LoC) — indexing gets slow.
- Languages without strong type info — the autocomplete leans heavily on types.
- Workflows that need to span outside the editor (use Claude Code or another shell agent).
Pros and cons
Pros
- Best-in-class tab autocomplete
- Familiar VS Code feel with all your existing extensions
- Strong agent mode with restore checkpoints
- Decent free tier to try before paying
Cons
- Pro tier required for serious daily use
- Indexing performance degrades on huge repos
- Some lock-in to Cursor's UI conventions
Real workflows using this tool
Prompt examples
Copy any of these, replace the placeholders, run.
Project rules file (.cursor/rules/general.mdc)
# Conventions for this codebase
- We use TypeScript strict mode. No any types in new code.
- Tests live next to source: foo.ts has foo.test.ts.
- Use functional React components with hooks. No class components.
- Format with prettier on save.
- Commit messages: type(scope): summary
- Types: feat, fix, chore, docs, refactor, test
# Don't
- Don't add new dependencies without asking.
- Don't refactor code outside the scope of the request.
- Don't suppress lint or type errors — fix them or surface them.Targeted edit with Cmd-K
Extract this component into its own file. Keep the same props API.
Update imports in this file but don't touch other files.Agent mode scoped task
Add a /pricing page using the same layout as /about.
Source the plan data from a new constants/pricing.ts file.
Run the build when you're done and fix any errors.
Don't touch /about or other existing pages.Alternatives
Frequently asked questions
Watch
Hand-picked videos from official + trusted channels. Opens in a new tab.
Related on AIKnowHub
Tool Guide
Claude Code Guide
The definitive guide to Anthropic's terminal-native AI coding agent — install, configure, MCP, hooks, skills, sub-agents, plan mode, cost, security, limitations.
Comparison
Best AI Model for Coding
Which model actually writes the best code? Benchmarked against real tasks — refactors, bug fixes, new features, code review.
Directory
Claude Code
Terminal-native agent that reads, writes, runs, and reviews your codebase.
Concept
AI Agents Explained
Agents are LLMs that can take actions in a loop. Here's what that actually means, where they shine, and where they fall over.
Learning Path
AI Coding Roadmap
Become 3x more effective with AI in your daily coding loop. Aimed at working engineers.
Comparison
Best AI Coding IDE
Cursor vs Windsurf vs Claude Code vs Aider vs Continue vs Copilot. Which AI-native editor is best for what kind of work?