OpenAI Codex

Run OpenAI Codex via AnyRouter: configure base_url in config.toml and your API key in auth.json or with cc-switch, then use it in the terminal or desktop app.

Use Codex via the AnyRouter API

Installation

Download from the official site (macOS)

https://openai.com/codex/

Install via the command line

npm install -g @openai/codex

Environment configuration

Configure via config files

Edit the ~/.codex/config.toml config file and add the following:

profile = "AnyRouter"

[model_providers.AnyRouter]
name = "AnyRouter"
base_url = "https://anyrouter.win/v1"
personality = "pragmatic"
wire_api = "responses"

[profiles.AnyRouter]
model = "gpt-5.2"
model_provider = "AnyRouter"
model_reasoning_effort = "high"

Edit the ~/.codex/auth.json config file as follows:

{
  "OPENAI_API_KEY": "AnyRouter_API_KEY"
}

Configure via cc-switch

Run CC-Switch and add a provider.

Select "AnyRouter" from the preset list.

Enter your API key in the "API Key" field and click "Add" to save the settings.

Back on the home page, select "AnyRouter" in the provider list and click "Enable" to start using it.

Using Codex

In the terminal

Open a terminal, go to your project directory, and run the codex command.

cd /path/to/your/project
codex

Set permissions as needed.

Choose the model you want to use.

Type a natural-language prompt. If it responds normally, the configuration succeeded.

In the Codex desktop app

Open the Codex desktop app and choose a working directory.
Type a task in the input box. If it responds normally, the configuration succeeded.

Handy Command Reference

Help command

codex -h

Full command options

Usage
  $ codex [options] <prompt>

Options
  -h, --help                 Show help and exit
  -m, --model <model>        Specify the model to use (default: codex-mini-latest)
  -i, --image <path>         Path to a file containing image input
  -v, --view <rollout>       View a previously saved session
  -q, --quiet                Non-interactive mode; print only the assistant's final output
  -a, --approval-mode <mode> Override the approval policy: 'suggest', 'auto-edit', or 'full-auto'

  --auto-edit                Automatically approve file edits; still prompt to confirm commands
  --full-auto                Automatically approve edits and commands in the sandbox environment

  --no-project-doc           Do not automatically include the repo's 'codex.md' file
  --project-doc <file>       Include the specified Markdown file as context
  --full-stdout              Do not truncate the stdout/stderr of command output

Dangerous options
  --dangerously-auto-approve-everything
                             Skip all confirmation prompts and execute commands directly (no sandbox protection)
                             For temporary local testing environments only

Experimental options
  -f, --full-context         Launch in "full context" mode, loading the entire repo into context
                             and applying batch edits in a single pass
                             Only compatible with the --model flag

Examples
  $ codex "Write and run a Python program that prints ASCII art"
  $ codex -q "fix the build issue"

OpenAI Codex

On this page