Introduction

Cursor is a powerful AI programming assistant that is changing the traditional coding experience with its “chat-style” programming mode and efficient functional modules. This article presents 30 practical tips for using Cursor, from basic concepts to advanced operations, to enhance your AI programming efficiency.

Basic Concepts

01. Understand “Chat-Style” Programming
Cursor marks the arrival of “chat-style” programming. Compared to traditional programming modes, it has three core breakthroughs: writing code in “natural language,” iterating at the speed of judgment, and blurring the boundaries between product managers, designers, and programmers. This new paradigm shifts our focus from “how to write code” to “what problem to solve,” pushing AI to compel you to “think clearly and articulate clearly.”

02. Familiarize Yourself with Cursor’s Four Core Features
Cursor provides different capabilities in various scenarios, ranging from simple to complex: Tab, Inline chat, Ask, and Agent. Understanding the characteristics and applicable scenarios of these four functional modules is the foundation for using Cursor efficiently.

03. Master the Transition from “Thinking Clearly” to “Articulating Clearly”
AI is powerful, but it does not know what you are thinking. For effective communication, it is recommended to use structured expressions with sufficient context; the most direct structured expression is to describe requirements in Markdown format, which naturally segments the content, making it easier for AI to understand.

04. Learn to Break Down Problems and Validate Small Steps
Break complex problems into smaller, simpler ones and solve them step by step. During development, avoid generating thousands of lines of code at once for validation; instead, execute and validate tasks incrementally to better control code quality.

05. Understand MCP (Model Context Protocol)
MCP is the “universal connector” between AI and the outside world, giving AI eyes and arms. Its true value lies in unifying standards, eliminating the need to reinvent the wheel, allowing AI to have a larger context and significantly improving closed-loop operability.

Daily Operations

06. Terminal Conversations
No more struggling to remember Linux commands; simply use command+k to describe command-line operations in natural language. This feature is particularly useful for local development, allowing Cursor to operate the local terminal directly.

07. Generate Comments for Historical Code
Select code and use command+k to quickly generate comments for historical code, which is much faster than the Ask mode. This is especially useful for taking over someone else’s code or reviewing your earlier code.

08. One-Click Commit Message Generation
Say goodbye to the hassle of thinking about “what did I change in my code?” Cursor can generate standardized commit messages with one click, improving Git operation efficiency.

09. Quickly Visualize Project Architecture
When taking over a new project, use the Ask mode to organize the project architecture diagram, outputting text in Mermaid syntax. You can paste it into https://mermaid.live/ to quickly understand the project structure.

10. Use Notepad to Record Key Ideas
Use Notepad to record important context, and use @ to call it. Notepad serves as a good bridge between Ask and Agent modes, helping maintain coherent thoughts.

11. Use @Git to Identify Code Vulnerabilities
When encountering a code MR (Merge Request), compare it with the main code to check for issues. If problems arise after the MR, use the @Git feature to quickly locate them.

12. Use Checkpoint for One-Click Rollback
If AI modifies code incorrectly, you can use the checkpoint feature to quickly roll back to a previous stable version, avoiding the hassle of manual recovery.

13. Set Custom Prompts
Set your custom prompts in Cursor Rules to improve AI’s understanding of your needs. There are many prompt templates available online for you to find and customize.

14. Drag and Drop to Add Context
No need to search through directories one by one to add context; simply hold the target file in the directory and drag it into the dialog box. This significantly improves work efficiency.

15. Use @web to Access the Latest Information
Utilize the online feature to quickly obtain the latest information, solving various problems encountered during development, especially regarding new technologies or libraries.

Advanced Techniques

16. One Question, One Chat
Break down large module requirements into smaller questions and open a separate Chat for each new question. Long conversations may lead to AI memory confusion and longer response times, hindering review and management.

17. Use Composer for Multi-File Modifications
When it involves data coordination between modules (multiple code files need to work together), it is recommended to use Cursor’s Composer feature. Compared to Chat, Composer can analyze multiple files simultaneously, understand code context, and provide more reasonable modification suggestions.

18. Tell Cursor Not to Rush into Writing Code
Cursor tends to provide code directly; in the early stages of a project, it’s better to have divergent discussions first, allowing AI to help clarify unclear details. Clearly instruct AI to hold off execution until your thoughts are confirmed.

19. Guide AI to Ask Questions, Avoid Mindless Execution
Encourage AI to ask clarifying questions to confirm more details. Cursor defaults to trusting your judgment; if you are unsure of the solution, make sure AI asks you questions to avoid executing based on incorrect reasoning.

20. Emphasize Not Modifying Unrelated Code
Clearly define the scope in your requirements description, indicating which code can be modified and which cannot, to reduce the probability of AI making incorrect changes. Emphasize that you are a coding novice, prompting AI to generate more detailed comments in Chinese to help understand code logic.

21. Create .md Requirement Documentation
Establish .md requirement documentation to record project background, core logic, and implemented features. Each time a new feature is developed, have AI read the document first to ensure understanding of the context. Clearly instruct AI to read the requirements to avoid missing key content due to excessive @ references across multiple files.

22. Emphasize “Chain of Thought” to Enhance AI Reasoning
Use the “Chain of Thought” technique to encourage AI to engage in more rigorous logical thinking, applicable in scenarios like complex calculations, code analysis, and task planning, reducing AI’s vague reasoning.

23. Add Debugging Code to Help Locate Issues
When implementing complex features, instruct AI to add debugging code, paste the code into the editor, and check the actual execution results. If the results do not meet expectations, provide screenshots to AI to help quickly locate the problem.

24. Have Claude Provide Rich Responses to Aid Understanding
Guide Claude to explain vague concepts in a richer manner, using symbols and text arrangements to provide a more intuitive understanding of differences and enhance comprehension of complex concepts.

25. Use Project Rules
Abandon .cursorrules and switch to Project Rules. It supports setting different rules by file type, controlling AI tone and structure, and can sync through GitHub teams, making Cursor better understand your tech stack.

Share a versatile rule saved as a .mdc file for use in your project:

You are an advanced AI prompt engineer, specializing in transforming basic prompts into comprehensive, context-rich instructions that maximize AI capabilities. Your expertise lies in structuring prompts that yield highly specific, actionable, and valuable outputs.

Core Process:

1. Deep Prompt Analysis

Thoroughly analyze the user’s original prompt to extract explicit and implicit intentions.

2. Strategic Prompt Enhancement

Transform the original prompt by incorporating clear role definitions, contextual background, and precise instructions.

3. Domain-Specific Optimization

Incorporate domain-specific terminology and reference relevant methodologies.

4. Structural Engineering

Organize the enhanced prompt using a clear hierarchical structure.

5. Quality Assurance

Review the enhanced prompt against criteria for completeness, specificity, actionability, flexibility, and error prevention.