AI Coding Tool Setup and Reference (4/7) — Codex CLI Install: macOS, Linux, Windows, WSL2

Install paths, auth, sandbox modes, AGENTS.md, config.toml — all in one page, per official docs

ํ•ต์‹ฌ ์š”์•ฝ

  • Audience: You've used Claude Code but Codex is new to you, or you plan to run both.
  • What you'll get: Per official OpenAI Codex docs — install (npm/brew), per-platform verbatim commands, ChatGPT vs API key auth, AGENTS.md starter template, config.toml location, three sandbox modes.
  • Prerequisite: Terminal basics. Node-based install, so you'll need Node eventually.

1. Install methods at a glance

Official docs lead with npm; Homebrew is secondary.

Method Platforms When to use
npm (recommended) macOS, Linux, Windows, WSL2 Most personal dev setups
Homebrew macOS, Linux When you manage tools with Homebrew
codex exec (non-interactive) All CI / script automation

Unlike Claude Code, Codex's npm install is not deprecated — OpenAI's docs explicitly lead with it.

2. System requirements

  • OS: macOS / Linux / Windows
  • Windows supports both native PowerShell (Windows sandbox) and WSL2
  • Linux: x86_64, arm64
  • macOS: Apple Silicon, x86_64
  • Node.js: Official README doesn't pin a version. Latest LTS (Node 20+) recommended.
  • Network: Required for ChatGPT or API auth.

3. Install — verbatim commands

3.1 npm (recommended)

npm i -g @openai/codex

Upgrade:

npm i -g @openai/codex@latest

Avoid sudo npm install -g for permission/security reasons. Use nvm or fnm for a user-level Node environment.

3.2 Homebrew (macOS / Linux)

brew install --cask codex

3.3 Windows — native PowerShell

npm i -g @openai/codex

Install happens in PowerShell, and subsequent runs use Windows' native sandbox. Controlled by the windows.sandbox_mode option in config.toml.

3.4 Windows — WSL2

From Ubuntu/Debian WSL2, run the Linux install.

npm i -g @openai/codex

WSL2 uses the Linux sandbox. If your project uses Linux-only toolchains (Docker, ripgrep, etc.), WSL2 is the better fit.

4. Verify the install

codex --version

If the command isn't found, check that the global npm bin is in your PATH:

npm bin -g

5. Auth — two options

Official docs specify two methods.

Option For Billing Setup
ChatGPT account (recommended) Plus / Pro / Business / Edu / Enterprise Subscription (Codex included) Browser login on first run
API Key Platform credits users Usage-based Choose "Use an API Key" at login

ChatGPT Plus, Pro, Business, Edu, and Enterprise plans all include Codex access. The free ChatGPT plan does not.

On first codex, a prompt asks "Sign in with ChatGPT" or "Use an API Key." Credentials are then stored under ~/.codex/.

6. First session

From your project directory:

cd /path/to/your/project
codex

Interactive session starts. Essential commands:

Command Effect
codex Start interactive session
codex exec "task" One-off non-interactive run (CI-friendly)
codex init Initialize project (generate AGENTS.md)
codex --version Show version

Slash commands inside a session (essentials): - /model — switch model (GPT-5, GPT-5-Codex, etc.) - /config — settings - /exit or Ctrl+D — quit

Full list in the official CLI Reference.

7. AGENTS.md — the Codex project instructions file

Codex reads AGENTS.md before any work. The concept parallels Claude Code's CLAUDE.md, but loading rules differ.

7.1 File placement and priority

Per official docs:

  1. Global: ~/.codex/AGENTS.md or ~/.codex/AGENTS.override.md
  2. Project: Walk down from the Git root to the current working directory
  3. Merge order: Concatenate top-down. Files closer to the current directory override earlier guidance.
  4. At most one file per directory (AGENTS.override.md wins over AGENTS.md in the same dir)

7.2 Starter — ~/.codex/AGENTS.md (global)


## Testing
- Run `npm test` after modifying JavaScript files
- Run `pytest -x` after modifying Python

## Dependencies
- Prefer `pnpm` over `npm`
- Ask before adding a new production dependency

## Commits
- Conventional commits (feat:, fix:, chore:)
- One topic per PR

7.3 Project override — services/payments/AGENTS.override.md


- Use `make test-payments` (not `npm test`)
- Notify the security channel before rotating API keys
- Legal review required before collecting new PII

AGENTS.override.md replaces upstream rules (not concat).

7.4 Coexisting with Claude Code's CLAUDE.md

When both tools operate in the same repo, have CLAUDE.md import AGENTS.md to avoid duplication (Claude Code side):

@AGENTS.md

## Claude Code

Use plan mode for changes under `src/billing/`.

8. ~/.codex/config.toml basics

Shared config between CLI and IDE extension. Location: ~/.codex/config.toml.

Key categories:

model = "gpt-5-codex"

provider = "openai"

approval_policy = "on-request"  # or "on-failure", "never"

sandbox_mode = "workspace-write"  # read-only / workspace-write / danger-full-access

[windows]
sandbox_mode = "elevated"  # native Windows sandbox option

Advanced settings: Config Reference. Start with defaults, add only as needed.

9. Three sandbox modes

Codex controls filesystem and network access per sandbox mode.

Mode Read Write Network
read-only ✅ ❌ ❌
workspace-write (default) ✅ ✅ (within cwd) ❌ (default)
danger-full-access ✅ ✅ (unrestricted) ✅

9.1 Allow extra paths in workspace-write

Default: writes limited to cwd. For additional paths:

[sandbox]
writable_roots = ["/tmp/build", "/var/log/myapp"]

9.2 Sandbox vs approval policy

  • approval_policy: when Codex pauses for approval
  • on-request: whenever a command leaves the sandbox
  • on-failure: only when a sandboxed command fails
  • never: never asks (automation-only, risky)
  • sandbox_mode: what Codex is allowed to do

Orthogonal axes — workspace-write + on-request is the most common personal dev setup.

10. Native Windows vs WSL2 — decision

Situation Pick
Node/Python-heavy project, no Docker Windows native
Linux-only toolchain (Docker, ripgrep, GNU make) WSL2
Large repo where filesystem perf matters WSL2 (work in ext4)
Heavy PowerShell script ecosystem Windows native

WSL2 initial setup: install "Ubuntu" from Windows Store → wsl --install → run npm install inside that shell.

11. Coexistence with Claude Code

Common concerns when running both in the same repo.

Potential conflict Resolution
.claude/ vs .codex/ Different directories — no conflict
AGENTS.md vs CLAUDE.md Claude Code can @AGENTS.md import — single source
Sandbox modes have different names Codex uses workspace-write; Claude Code uses /sandbox — independent
/model command Both use the name but with their own model lists

Concrete Codex + Claude Code parallel workflows are covered in a later post.

12. Counter-scenarios — when this guide doesn't fit

  • Free ChatGPT plan only → No Codex access. Plus-or-higher, or API key, required.
  • macOS 12 or older → Official support is on recent macOS. Upgrade OS first.
  • You don't want a Node install → Official path is npm-centric. Homebrew is the alternative but doesn't auto-update.
  • Company policy blocks writes outside /tmp → Start with sandbox_mode = "read-only", then approve required commands via the prompt.

13. What's next

Once installed:

  1. ~/.codex/config.toml advanced configuration — approval / sandbox / model switching (coming soon).
  2. Codex + Claude Code parallel workflow — role division (coming soon).
  3. Claude Code install complete guide — for comparison or parallel use.

References

This is post 6/15 in the "AI Coding CLI Entry Guide" series. last verified: 2026-04-25 (per official OpenAI Codex CLI docs + GitHub README).

Series overview: Series index

๋Œ“๊ธ€

๋Œ“๊ธ€ ์“ฐ๊ธฐ

์ด ๋ธ”๋กœ๊ทธ์˜ ์ธ๊ธฐ ๊ฒŒ์‹œ๋ฌผ

Agent Memory Engine (2/10) — Building an AI Agent Memory System with SQLite Alone

← 1/10 SQLite Memory Engine f… ๐Ÿ“š Series Index 3/10 memcore Completed → A memory engine that runs without daemons, without dependencies, anywhere ํ•ต์‹ฌ ์š”์•ฝ This post covers how to design an AI agent memory engine (memcore) on a single-file SQLite backend. - Structure: 25 modules, ~6,464 lines, 22 DB tables - Principles: no daemon, single-file portability, no required external dependencies, host-agnostic - Core techniques: 4-layer hybrid search (topic · vector · FTS5 · LIKE), U-tag dialectic-based confidence evolution, ingestion-layer bias rejection rules What This Post Covers One approach to building AI agent memory without a server or cloud vector DB. Specifically: (1) the limitations of text-file-based memory, (2) vector DB alternatives compared and selection criteria, (3) how to layer search tiers, (4) how to evolve confidence scores over time, and (5) how to block biased memory accumulation at the ingestion stage. Design Principles: Why Single-File SQLite Text-file-based ...
Read more »

"ML Foundations (9/9) — PyTorch vs TensorFlow, and the Road to Local LLMs"

์ด๋ฏธ์ง€
← 8/9 Deep Learning Architec… ๐Ÿ“š Series Index (series end) The final part. So far we've focused on models. Now we focus on the tools that actually run them . We'll lay out the philosophical difference between PyTorch and TensorFlow, and trace how Hugging Face transformers , llama.cpp , MLX , and Ollama built the bridge to running large language models on your own machine. By the end you should have the full mental model of "download a pretrained LLM and serve it locally." 0. Learning Objectives Explain the eager-vs-graph difference between PyTorch and TensorFlow. Explain, in graph terms, how autograd automates the backward pass. Use Hugging Face transformers ' AutoModel , AutoTokenizer , and pipeline abstractions. Describe llama.cpp's GGUF quantization, INT4 inference, and CPU-first flow. Describe Apple MLX's use of unified memory and how it differs from PyTorch. Run a local LLM with Ollama and call it through an OpenAI-compatible API. ...
Read more »

"RAG Core Study (14/26) — Evaluation Sets with RAGAS & DeepEval"

์ด๋ฏธ์ง€
Series overview: Series index ๐Ÿ“š Series Index Without an evaluation set, every RAG improvement is a story, not evidence. RAG systems fail in more than one place: retrieval, context selection, answer generation, and grounding. That means "looks better" is not a metric. Part 14 explains how to build a golden dataset , how RAGAS and DeepEval fit into the workflow, and why teams must separate retrieval quality from answer quality if they want tuning decisions to hold up. 0. Prerequisites Part 13 reranking — system quality now depends on multi-stage ranking. Part 1 grounding — a correct-looking answer is not enough. Part 12 Hybrid — multiple retrieval settings need comparison on fixed data. 1. Learning Objectives Build a small but useful golden dataset for RAG. Distinguish retrieval metrics from answer metrics . Use RAGAS and DeepEval in the right roles. Avoid the common traps in judge-model-based evaluation. 2. ํ•ต์‹ฌ ์š”์•ฝ An evaluation set for RAG is ...
Read more »

"ML Foundations (8/9) — Deep Learning Architectures: CNN, RNN, Attention"

์ด๋ฏธ์ง€
← 7/9 Deep Learning Training ๐Ÿ“š Series Index 9/9 PyTorch vs TensorFlow → What did we keep adding on top of an MLP? Vision went CNN, sequences went RNN/LSTM, and eventually both converged on Attention and the Transformer. Each architecture is easier to remember if you read it as what it gave up to gain something else . This part is the one-line summary of the decisive inflection points: LeNet → AlexNet → ResNet → LSTM → Attention → Transformer. 0. Learning Objectives Explain why CNN's convolution, pooling, and weight sharing match images so well. Trace what got unblocked from LeNet → AlexNet → VGG → ResNet as depth grew. State BPTT for RNNs and the vanishing/exploding gradient problem. Write the LSTM gate equations (forget, input, output) with intuition. Write the attention formula in query/key/value form and the scaled dot-product variant. State the precise reasons Transformers replaced RNNs (parallelism + long-range dependencies). 1. ํ•ต์‹ฌ ์š”์•ฝ CNN : weight sharin...
Read more »

"ML Foundations (7/9) — Deep Learning Training: Optimizers, Regularization, Initialization"

← 6/9 Neural Networks ๐Ÿ“š Series Index 8/9 Deep Learning Architec… → Building an MLP, as we did in Part 6, is not the same as getting it to train well . The right optimizer, regularization, initialization, and learning rate — when these line up, deep networks converge. When they don't, the network refuses to learn at all. This part is the catalog of those crafts, with formulas, paper citations, and code in one place. 0. Learning Objectives Compare and write the update rules for SGD, Momentum, Nesterov, and Adam. Explain how Dropout, BatchNorm, and LayerNorm work and where they belong in a model. Derive the variance formulas for Xavier and He initialization and match them to activations. Implement step, cosine, and warmup learning-rate schedules in PyTorch. Explain why gradient clipping is effectively required for RNNs and Transformers. Diagnose the most common training failures (NaN, plateau, overfitting) and apply first-line fixes. 1. ํ•ต์‹ฌ ์š”์•ฝ SGD : \(w \leftarro...
Read more »

OpenClaw to Hermes Migration (2/13) — What to Preserve, Partially Port, or Discard

← 1/13 Current Structure Inve… ๐Ÿ“š Series Index 3/13 What Moves to Hermes → A code review record classifying operational scripts into three tiers: preserve, partial-preserve, and discard. What This Post Covers An asset classification framework (preserve / partial-preserve / discard) for migrating a legacy agent system to a successor A concrete case applying the Strangler Fig + Subset Migration strategy to a memory pipeline Classification results evaluated across LOC, external dependencies, and replaceability axes A method for slimming down a bloated single file (1,310 LOC) using the migration as a natural entry point Problem Definition: Port Everything or Cut Deep? The OpenClaw inventory comprises multiple agents, multiple scheduler/daemon processes, and multiple operational scripts. When moving to the successor system, Hermes, the choice is not binary. A full port carries accumulated technical debt along with it; a minimal port discards hard-won domain logic. Code-revie...
Read more »

AI Agents I Built (5/7) — Building an Automated Blogger API Publishing System

← 4/7 My Life Organizer Agent ๐Ÿ“š Series Index 6/7 Dual → Designing a Markdown-Based Auto-Publish Pipeline with Blogger API v3 + OAuth 2.0 Key Summary Blogger API v3 + OAuth 2.0 enables a pipeline that converts Markdown files into publishable HTML posts blogger_publish.py handles the full flow — frontmatter parsing → body cleanup → HTML conversion → CSS injection → API post — in a single script For bulk publishing, quota management (delay, retry, TOC suppression) is the critical design concern Platform Selection Criteria API access is mandatory for AI agent-driven content publishing. A comparison of platforms: Platform API Cost AdSense Decision WordPress.com REST API Paid plan required Paid only Cost overhead Ghost REST API Self-hosted or paid Manual setup Ops overhead Blogger v3 API Free Native integration Adopted Blogger is free, exposes a REST API, and integrates natively with AdSense. For personal blog automation, this combination i...
Read more »