OpenAI Codex CLI — Official Coding Agent

What it is

OpenAI Codex CLI is OpenAI's official open-source coding agent that runs in the terminal. It connects to GPT models to generate code, edit files, run commands, and scaffold projects. The CLI supports custom prompts, agent skills for specialized tasks, and integration with code editors.

Codex CLI targets developers who prefer terminal-based workflows over GUI-based AI coding tools. It provides the same code generation capabilities as cloud-based solutions but runs locally in your terminal with direct file system access.

How it saves time or tokens

Codex CLI operates directly on your codebase. It reads project files, understands context, and makes targeted edits without you needing to copy-paste code into a chat interface. The agent maintains conversation context across multiple interactions, so you can iteratively refine code without re-explaining the project structure.

Custom agent skills let you define reusable workflows for common tasks like generating tests, refactoring modules, or creating documentation. Once configured, these skills execute with a single command.

How to use

1. Install Codex CLI:

npm install -g @openai/codex

1. Set your API key:

export OPENAI_API_KEY=your-key-here

1. Start Codex in your project directory:

codex

1. Or run without installing:

npx @openai/codex

Example

# Generate a REST API handler
codex 'Create a POST /api/users endpoint that validates
email and password, hashes the password with bcrypt,
and saves to the users table'

# Run with a specific model
codex --model gpt-4o 'Refactor this file to use async/await
instead of callbacks'

# Use in non-interactive mode for CI/CD
codex --quiet --approve 'Add JSDoc comments to all
exported functions in src/utils/'

Codex reads the surrounding code, generates the changes, and presents a diff for review before applying.

Related on TokRepo

• AI coding tools -- Compare other AI-powered development tools
• Automation tools -- Browse workflow automation for developers

Common pitfalls

• Codex CLI requires an OpenAI API key with access to GPT-4 or newer models. The agent consumes tokens for every interaction, and complex tasks can use significant token budgets. Monitor your API usage.
• The CLI operates on files in the current directory. Running it in the wrong directory can result in edits to unintended files. Always verify your working directory before starting a session.
• Agent skills are defined in configuration files. Poorly written skill prompts produce inconsistent results. Test skills on small files before applying them to entire projects.