Codex CLI can use the OpenAI-compatible endpoint of cheapkeyai.com. Configuration should be placed in the user's `.codex` directory for multi-project usage.
Automatic Installation
curl -fsSL https://cheapkeyai.com/cdn/codex/setup.sh | bash
irm https://cheapkeyai.com/cdn/codex/setup.ps1 | iex
Install Codex
npm i -g @openai/codex@latest
Create config.toml
model_provider = "cheapkeyai"
model = "gpt-5.2"
model_reasoning_effort = "xhigh"
model_verbosity = "high"
network_access = "enabled"
disable_response_storage = true
windows_wsl_setup_acknowledged = true
[model_providers.cheapkeyai]
name = "cheapkeyai"
base_url = "https://cheapkeyai.com/v1"
wire_api = "responses"
requires_openai_auth = true
Create auth.json
{
"OPENAI_API_KEY": "sk-your-api-key-here"
}
Usage
codex
codex -m gpt-5.3-codex -c model_reasoning_effort="xhigh"
codex "Review current changes and suggest tests to run"
Troubleshooting
- Codex not connecting: check `base_url`, `wire_api`, and `auth.json`.
- Command not found: check if npm global bin is in PATH.
- Weird model output: try another model or adjust reasoning effort settings.
- Cannot access network: enable network configuration if web/API access is needed.