# How to Install the OpenAI Codex CLI: 4 Methods from npm to Source > Install the OpenAI Codex CLI with four methods: npm, Homebrew, direct download, or build from Rust source. Get started quickly with our comprehensive guide. - Repository: [OpenAI/codex](https://github.com/openai/codex) - Tags: how-to-guide - Published: 2026-03-06 --- **You can install the OpenAI Codex CLI via npm for legacy TypeScript support, Homebrew for native macOS binaries, direct GitHub downloads for Linux/Windows, or by building the Rust source code directly.** The OpenAI Codex CLI is a command-line interface for the `openai/codex` repository that enables AI-powered coding assistance directly in your terminal. Whether you need the legacy TypeScript implementation or the modern Rust-based binary with full sandboxing capabilities, this guide covers every installation method with specific commands and source file references. ## Installation Methods Overview The Codex CLI offers four distinct installation paths, each suited to different operating systems and workflow preferences: - **npm (legacy TypeScript CLI)** – Cross-platform Node.js package with a JavaScript shim wrapping the Rust binary - **Homebrew cask (macOS)** – Pre-built native binary managed through Homebrew - **Direct download** – Platform-specific binaries from GitHub Releases for Linux and Windows - **Build from source** – Compile the Rust implementation locally for development or custom builds ## Method 1: Install via npm (Legacy TypeScript CLI) The npm distribution packages the CLI as `@openai/codex` and works on any platform running Node.js 16 or higher. ### Installation Steps Install the package globally to access the `codex` command from anywhere: ```bash npm i -g @openai/codex ``` Verify the installation succeeded: ```bash codex --version ``` ### Architecture Notes According to the `openai/codex` source code, the npm package contains a thin JavaScript shim located in the legacy `codex-cli/` directory. This shim launches the compiled Rust binary that ships as a bundled asset within the npm package. The actual CLI implementation resides in [`codex-rs/cli/src/main.rs`](https://github.com/openai/codex/blob/main/codex-rs/cli/src/main.rs), which handles subcommand dispatching. ## Method 2: Install via Homebrew on macOS For macOS users seeking a zero-dependency installation, Homebrew provides the Rust-based native binary through its cask system. ### Installation Steps Install the pre-built binary using Homebrew's cask command: ```bash brew install --cask codex ``` Confirm the binary is available: ```bash codex --version ``` ### Why Use Homebrew? The Homebrew distribution delivers the Rust-based implementation described in [`codex-rs/README.md`](https://github.com/openai/codex/blob/main/codex-rs/README.md), which is the canonical and most feature-complete version of the CLI. This method avoids requiring Node.js or Rust toolchains on your system, as the cask downloads a pre-compiled binary that includes all subcommands (`exec`, `tui`, and `cli`). ## Method 3: Direct Download from GitHub Releases Linux and Windows users, or those who prefer manual installation, can download platform-specific binaries directly from the repository's releases page. ### Linux and Windows Installation 1. Navigate to the GitHub Releases page: `https://github.com/openai/codex/releases` 2. Download the asset matching your platform (e.g., `codex-linux-x86_64.tar.gz` or `codex-windows-x86_64.zip`) 3. Extract the archive and move the binary to a directory in your `$PATH` For Linux systems: ```bash tar -xzf codex-linux-x86_64.tar.gz -C ~/.local/bin chmod +x ~/.local/bin/codex codex --version ``` This method provides the same Rust-based binary found in the Homebrew distribution, including the sandboxing capabilities documented in [`codex-rs/linux-sandbox/README.md`](https://github.com/openai/codex/blob/main/codex-rs/linux-sandbox/README.md). ## Method 4: Build from Source (Rust) Developers who need the latest unreleased changes or want to contribute to the project can compile the CLI from the Rust source code. ### Prerequisites Install the required Rust toolchain and task runner: ```bash cargo install just cargo install --locked cargo-nextest # Optional, for testing ``` ### Building the Binary Clone the repository and build the project: ```bash git clone https://github.com/openai/codex.git cd codex just build ``` The `just build` command compiles the `codex` binary using the configuration defined in [`codex-rs/cli/Cargo.toml`](https://github.com/openai/codex/blob/main/codex-rs/cli/Cargo.toml). The resulting binary appears in `target/release/codex`. Install the binary to your local bin directory: ```bash cp target/release/codex ~/.local/bin/ codex --version ``` ### Source Code Structure The Rust implementation in `codex-rs/` represents the canonical CLI architecture: - **[`codex-rs/cli/src/main.rs`](https://github.com/openai/codex/blob/main/codex-rs/cli/src/main.rs)** – Entry point that parses arguments and dispatches to subcommands - **[`codex-rs/cli/Cargo.toml`](https://github.com/openai/codex/blob/main/codex-rs/cli/Cargo.toml)** – Defines the binary target and dependencies - **`codex-rs/core/src/config_loader`** – Handles configuration file loading from `~/.codex/` directories - **`codex-rs/linux-sandbox/`** – Implements the sandboxing policies accessible via the `--sandbox` flag ## Verifying Your Installation Regardless of installation method, confirm the CLI functions correctly by checking its version and running a test command: ```bash codex --version codex chat "Write a Python function that returns fibonacci numbers" ``` The binary should respond with version information and process the natural language request using the OpenAI API. ## Summary - **npm** provides the quickest cross-platform installation using `@openai/codex`, wrapping the Rust binary in a Node.js shim. - **Homebrew** delivers a native macOS binary without external dependencies via `brew install --cask codex`. - **Direct download** from GitHub Releases suits Linux and Windows users who want pre-built binaries without package managers. - **Building from source** requires the Rust toolchain and the `just` task runner, compiling the canonical implementation found in [`codex-rs/cli/src/main.rs`](https://github.com/openai/codex/blob/main/codex-rs/cli/src/main.rs). All methods ultimately install the same multitool binary supporting `exec`, `tui`, and `cli` subcommands with full sandboxing capabilities. ## Frequently Asked Questions ### What is the difference between the npm and Homebrew installations? The npm package (`@openai/codex`) installs a legacy TypeScript wrapper that launches the Rust binary, requiring Node.js 16 or higher. The Homebrew cask installs the native Rust binary directly to your macOS system without Node.js dependencies. Both provide the same `codex` command, but Homebrew offers better performance and the canonical implementation as described in [`codex-rs/README.md`](https://github.com/openai/codex/blob/main/codex-rs/README.md). ### Do I need Node.js installed to use the Codex CLI? You only need Node.js if you install via npm. The Homebrew, direct download, and source build methods provide native binaries that do not require Node.js or any JavaScript runtime. The Rust-based implementation in `codex-rs/cli/` compiles to a standalone executable. ### How do I update the Codex CLI to the latest version? Update methods depend on your installation path. For npm, run `npm update -g @openai/codex`. For Homebrew, use `brew upgrade --cask codex`. If you built from source, pull the latest changes with `git pull` and rerun `just build` in the repository root, then copy the new binary from `target/release/codex`. ### Where are configuration files stored after installation? The CLI loads configuration from multiple sources as implemented in `codex-rs/core/src/config_loader`. It reads environment variables, `.env` files in the working directory, and user-specific settings stored in the `~/.codex/` directory. The binary also respects CLI flags like `--sandbox` which override config file settings without modifying global configuration.