NoxGate

Codex

Use an OpenAI Responses-compatible custom Model Provider.

Before you connect

  • Install and launch the tool.
  • Create an API key and save the plaintext value shown once.
  • Choose a model ID whose protocol and capability match the tool.

Minimal configuration

Register NoxGate in ~/.codex/config.toml and provide the key through an environment variable.

TOML + shell
# ~/.codex/config.toml
model = "YOUR_MODEL_ID"
model_provider = "nox_gate"

[model_providers.nox_gate]
name = "NoxGate"
base_url = "https://api.noxgate.net/v1"
env_key = "NOX_GATE_API_KEY"
wire_api = "responses"

# 启动 Codex 的 shell
export NOX_GATE_API_KEY=sk-your-api-key
codex
  • Persist the variable in your shell on macOS/Linux or as a user environment variable on Windows.

Verify the connection

  1. Save the configuration and restart Codex.
  2. Send a minimal prompt and confirm the Responses request succeeds.

Limitations and common issues

  • base_url must include /v1 and wire_api must be responses.
  • A 401 usually means the launched process cannot read NOX_GATE_API_KEY.