AI / Core OpenAI Codex Application Fundamentals Interview Questions
What is the Codex CLI and what are its key features?
The Codex CLI is an open-source, terminal-native agentic coding tool that brings OpenAI's coding models directly into your command line. It was rebuilt in 2025 around agentic workflows, making it significantly more capable than a simple code-generation prompt wrapper.
| Feature | Description |
|---|---|
| Agentic execution | Works through multi-step tasks autonomously - reads files, writes code, runs tests |
| Image input | Attach screenshots, wireframes, and diagrams to build shared context |
| To-do tracking | Tracks progress on complex tasks with a visible to-do list |
| Web search | Built-in web search for looking up APIs and documentation |
| MCP support | Connect to Model Context Protocol servers for external tool integration |
| Three approval modes | read-only, auto (workspace-scoped), full-access (network + anywhere) |
| Model selection | --model flag or /model command; defaults to gpt-5.4 if unspecified |
| Thread management | /model command switches model mid-thread; --model flag sets it at start |
# Installation npm install -g @openai/codex # Basic usage - interactive session codex # Run with a specific model codex --model gpt-5.5 # Execute a single task non-interactively codex exec "Add comprehensive docstrings to all public functions in src/" # Exec with specific model codex exec --model gpt-5.4-mini "Fix all linting errors in utils.py" # Change model during an active thread: # /model gpt-5.5 # Specify model in config.toml: # [codex] # model = "gpt-5.5"
The CLI and IDE extension share the same config.toml configuration file. Authentication supports both ChatGPT account sign-in (for ChatGPT subscribers) and API key authentication (for direct API access).
What are the three approval modes available in the Codex CLI?
Which configuration file do the Codex CLI and IDE extension share?
More Related questions...
