This chapter reveals why the “CLI versus IDE” debate is a false dilemma, and you’ll discover how professional teams assemble a multi-tool workday: Qt Creator for builds and debugging, Gemini CLI for automation and SSH, Cursor—for deep understanding of Qt architecture.
We’ll show refactoring of a real project from one main.cpp into a modular structure, where one tool completes the task in ~10 minutes, and another in ~5 minutes. Plus—3 Cursor modes (Agent/Ask/Plan), context management through @/#, and a brief review of the CLI/IDE player ecosystem.
If you’re currently using AI “as the mood strikes,” it’s better not to skip this chapter—otherwise you’ll keep paying with time where others have long been paying with proper process setup.
This chapter includes ready-to-use code and prompt examples.
Chapter Self-Assessment
How do CLI tools fundamentally differ from AI-IDEs in the context of Qt development?Answer
Correct answer: CLI excels in scripts, automation, and SSH sessions on servers, while AI-IDEs provide a visual environment with deep integration where AI understands the entire project architecture, including Qt signals/slots, classes, and dependencies.
Why has parallel use of Qt Creator together with Cursor or CLI utilities become a recognized industry practice?Answer
Correct answer: This allows effective distribution of functional tasks: Qt Creator serves as the central platform for building, debugging, and project lifecycle management, while specialized editors provide enhanced navigation, refactoring, and AI assistant interaction capabilities.
What is MCP (Model Context Protocol) and what capabilities does it open for Qt developers?Answer
Correct answer: MCP is an open protocol from Anthropic allowing AI models to interact with external tools and data. For Qt, this means the ability to read project structure, execute build commands (qmake, cmake), work with databases, read Figma layouts, and even analyze console logs during debugging.
Why must Gemini CLI be launched specifically from the project root?Answer
Correct answer: For Gemini CLI to see the entire structure and files of the project, it must be launched from the root directory, from where it can access all files through @ symbols and properly understand the Qt application architecture.
What’s the key difference between Cursor’s three modes: Agent, Ask, and Plan?Answer
Correct answer: Agent mode automatically executes multi-step tasks with changes across multiple files; Ask mode only analyzes and answers questions without changing code; Plan mode creates a detailed step-by-step development plan before executing major tasks.
Why set up Project Rules in Cursor, and how do they affect generated code quality?Answer
Correct answer: Project rules teach AI your coding style, naming conventions, architectural patterns, and Qt project specifics, making generated code look consistent, match team internal standards, and not violate app architecture.
In which practical scenarios are CLI tools preferable to full-fledged AI-IDEs?Answer
Correct answer: CLI is ideal for working on remote servers via SSH, automating builds in CI/CD pipelines, writing scripts for routine operations, and situations where maximum launch speed without GUI is needed.
Why is iterative development with AI more effective than trying to get perfect code in one request?Answer
Correct answer: A phased approach allows starting with a minimal working implementation, ensuring compilation, then sequentially improving code with short specific prompts (error handling, thread safety, logging), maintaining full control and not breaking architecture.
Which Qt-specific aspects are critically important to check after AI code generation?Answer
Correct answer: You must check: correct parent ownership (each QObject has a parent or is deleted via deleteLater), proper thread handling (moveToThread, Qt::QueuedConnection), use of modern signal/slot syntax, and presence of Q_OBJECT macro in all classes with meta-object system.
How are the @, #, and / symbols used in Cursor for context management?Answer
Correct answer: @ is used to reference code elements (classes) and web links; # to add specific project files to context; / to invoke built-in commands (e.g., /test to generate unit tests).
Which Cursor mode should be used for adding major functionality like integration with a new API or migration to Qt6?Answer
Correct answer: Plan mode, which turns an idea into a structured step-by-step development plan, allowing time savings and reducing the risk of breaking the project when adding large-scale changes.
Why is using screenshots and visual references more effective than text descriptions when creating UI with AI?Answer
Correct answer: AI models understand images and can generate QML or QtWidgets markup directly from visual references, automatically determining colors, sizes, spacing, and element layout, which is more accurate and faster than verbose text descriptions.
How did Cursor outperform Gemini CLI in the practical audio player refactoring comparison?Answer
Correct answer: Cursor demonstrated better understanding of Qt project structure, suggested more detailed breakdown into logically independent classes, and completed the task faster (in 5 minutes versus 10 minutes) thanks to tighter integration with sources and project context.
What potential security risks are associated with using MCP servers, and how can they be minimized?Answer
Correct answer: MCP servers get direct access to system files and commands. Risks are minimized by using only verified modules, prohibiting access to production secrets (passwords, tokens, keys), and pre-approving the server list with the security department for corporate projects.
Practical Assignments
Easy Level
Setting Up Gemini CLI and First Refactoring
Install Gemini CLI on your computer, go through authorization, and create a simple Qt application with one main.cpp file containing a class with QLabel and QPushButton. Use Gemini CLI to split the code into separate header and implementation files (.h and .cpp).
Hints: Install via npm install -g @google/gemini-cli. Launch gemini from the project root. Use @ symbol to add main.cpp to context. Formulate a prompt following the pattern of listing 69.1 specifying Qt and C++ version, refactoring goal, and file structure requirements.
Medium Level
Cursor: Creating SettingsManager with Project Rules
Install Cursor and set up project rules for Qt/C++ development (using examples from listings 69.2 and 69.3). Then in Agent mode, create a SettingsManager class with JSON settings persistence, change notification signals, and thread-safe implementation. Verify that generated code follows the specified style rules.
Hints: Open Cursor Settings via Ctrl+Shift+J. Add general rules to User Rules and C++ rules to Project Rules. Use Agent mode (Ctrl+.) for multi-step tasks. Check for Q_OBJECT presence, correct parent ownership, modern signal syntax. Use an iterative approach: first basic implementation, then refinements.
Hard Level
MCP Server Integration and Visual UI Refactoring
Connect an MCP server for file system or Git to Cursor. Create a Qt Widgets application with multiple windows. Then draw on paper or in a graphics editor an improved interface layout for one of the windows, upload a screenshot to Cursor, and ask AI to refactor the UI according to the visual reference. Ensure the MCP server is used for automatic Git change commits.
Hints: Explore the MCP server catalog at mcpservers.org. Configure MCP in Cursor Settings. Use @ symbol to attach the layout image and # to specify specific .ui or .cpp files. In the prompt, indicate: “Refactor UI according to the layout, preserve functionality, update layout and styles”. Use Plan mode to create a plan, then Agent for execution. Don’t forget to verify MCP server security before use.
💬 Join the Discussion!
Tried Gemini CLI or Cursor in your Qt projects? Discovered interesting MCP server use cases?
Maybe you have questions about choosing between CLI and AI-IDEs, setting up project rules, or integrating AI assistants into your team workflow?
Share your refactoring automation experience, talk about unexpected results from working with agent modes, or help other readers understand the practical aspects of AI development with Qt!