Codex CLI

Configure OpenAI Codex CLI to use the Model Gate Responses endpoint.

No current prepared translation is available. The English source is shown without machine translation.

Codex CLI through Model Gate

Codex uses the OpenAI Responses protocol. Its provider base_url includes /v1.

1. Install Codex

npm install -g @openai/codex
codex --version

Check the current official installation instructions at https://developers.openai.com/codex/.

2. Configure the provider

Edit:

Linux or macOS: ~/.codex/config.toml
Windows: %USERPROFILE%\.codex\config.toml

model = "gpt-5.4"
model_provider = "model_gate"
model_reasoning_effort = "high"

[model_providers.model_gate]
name = "Model Gate"
base_url = "https://api.model-gate.com/v1"
wire_api = "responses"
env_key = "MODEL_GATE_API_KEY"

wire_api = "responses" sends requests to /v1/responses.

3. Set the API key

Linux or macOS

export MODEL_GATE_API_KEY="mg_live_..."

To make it permanent, add the export to your shell profile or secret manager.

Windows PowerShell

[Environment]::SetEnvironmentVariable(
  "MODEL_GATE_API_KEY",
  "mg_live_...",
  "User"
)

Open a new terminal after setting a persistent variable.

4. Start Codex

codex

Reasoning effort

model_reasoning_effort = "low"

low: faster and shorter.
medium: balanced.
high: more thorough for complex work.

The upstream model must support the requested reasoning controls.

Troubleshooting

Problem	Resolution
Authentication error	Confirm `MODEL_GATE_API_KEY` is visible in the same terminal.
Codex uses OpenAI directly	Confirm `model_provider = "model_gate"` and the provider section name matches.
Unsupported endpoint	Confirm `wire_api = "responses"` and `base_url` ends with `/v1`.
Model not found	Query `/v1/models` for a canonical ID, or use an enabled alias confirmed by your administrator.

Do not set requires_openai_auth = true for this provider: that option selects OpenAI account authentication instead of the custom env_key.