# How to Run Codex from the Command Line

> Run Codex from the command line with codex exec: install the Codex CLI, run a one-shot prompt, pipe files, save answers and get JSON output.

You can run Codex from the command line without opening the chat screen. The `codex exec` command runs Codex in non-interactive mode: it takes one prompt, runs the task, prints the answer and exits. Use it in scripts, cron jobs or for a quick question.

## Install Codex CLI

On macOS or Linux, run the installer:

```bash
curl -fsSL https://chatgpt.com/codex/install.sh | sh
```

On Windows, run this in PowerShell:

```powershell
powershell -ExecutionPolicy ByPass -c "irm https://chatgpt.com/codex/install.ps1 | iex"
```

You can also install it with npm:

```bash
npm install -g @openai/codex
```

Or with Homebrew:

```bash
brew install --cask codex
```

Then start the Codex command line once and choose **Sign in with ChatGPT**. You can also use an OpenAI API key.

```bash
codex
```

## Run a prompt with codex exec

Put the prompt in quotes after `codex exec`:

```bash
codex exec "what is the purpose of life? answer one word"
```

Here is what the Codex CLI output looks like:

![Codex exec running a one-word prompt in the terminal and printing the answer Love](/assets/en/blog/how-to-run-codex-from-command-line/codex-exec-run-from-command-line.webp)

Codex prints:

- a header with the version, working folder, model, sandbox, approval mode and session id;
- your prompt under `user`;
- the answer under `codex`;
- the number of tokens used.

Then it exits and returns you to the shell.

Interactive session or one-shot run:

![Diagram: need a back-and-forth chat? Run codex. Otherwise run codex exec with a prompt](/assets/en/blog/how-to-run-codex-from-command-line/diagram_task-codex-need.png)

## Useful codex exec options for scripts

**Send a file or command output to Codex.** Piped text is added to your prompt:

```bash
cat error.log | codex exec "explain this error in two sentences"
```

**Read the whole prompt from a file:**

```bash
cat prompt.txt | codex exec -
```

**Save the final answer to a file.** Codex still prints it in the terminal:

```bash
codex exec "summarize README.md in three bullets" -o answer.txt
```

How input and output flow:

![Diagram: a prompt and piped text go into codex exec, and the answer goes to the terminal, an answer file or JSON](/assets/en/blog/how-to-run-codex-from-command-line/diagram_prompt-piped-text.png)

**Get JSON for a script.** Each event comes as one JSON line:

```bash
codex exec --json "summarize the repo structure"
```

**Let Codex change files.** By default `codex exec` runs read-only, unless `~/.codex/config.toml` sets another sandbox (as in the screenshot). To allow edits in the current folder:

```bash
codex exec --sandbox workspace-write "add a .gitignore for a Node.js project"
```

**Run outside a Git repository.** Codex stops outside a Git repo. Skip the check if you trust the folder:

```bash
codex exec --skip-git-repo-check "what is the purpose of life? answer one word"
```

**Continue the last run** with a follow-up prompt:

```bash
codex exec resume --last "now explain it in one sentence"
```

For all flags, run `codex exec --help` or see the [Codex non-interactive mode docs](https://developers.openai.com/codex/noninteractive).

## Related guides

- [How to Run Claude Code from the Command Line](/blog/how-to-run-claude-code-from-command-line/), the same task with `claude -p`.
- [Codex YOLO Mode: Run Without Approvals Safely](/blog/codex-yolo-mode-run-without-approvals-safely/), before you turn off approval prompts.

## Frequently Asked Questions

### How do I choose the model for codex exec?

Add `-m` (or `--model`) with the model name, for example `codex exec -m <model> "your prompt"`. Without it, Codex uses the model from your config file. The header of each run shows which model it used.

---

[View the canonical page](https://revdoku.com/blog/how-to-run-codex-from-command-line/) · [Browse llms.txt](https://revdoku.com/llms.txt)