Codex Configuration Guide

This guide shows how to configure Codex to use the OmniRouters OpenAI Compatible API.

1. Create the Codex configuration file

Linux / macOS:

mkdir -p ~/.codex
vim ~/.codex/config.toml

Windows:

C:\Users\YourUsername\.codex\config.toml

If the .codex directory or config.toml file does not exist, create it manually.

2. Write the API configuration

Add the following content to config.toml, and replace YOUR_API_KEY with your OmniRouters API key.

model = "gpt-5.5"
model_provider = "thirdparty"

[model_providers.thirdparty]
name = "My OpenAI Compatible API"

# API base URL
api_base_url = "https://omnirouters.com/v1"

# Whether to use official OpenAI authentication
requires_openai_auth = false

# Your API key
api_key = "YOUR_API_KEY"

# Use the OpenAI Responses API
wire_api = "responses"

# Disable WebSocket
supports_websockets = false

Field reference:

Field	Description
model	Default model used by Codex, for example gpt-5.5.
model_provider	Provider name used by the model. It must match thirdparty in [model_providers.thirdparty].
api_base_url	OmniRouters API base URL. For Codex, use https://omnirouters.com/v1.
requires_openai_auth	Set to false so Codex does not use official OpenAI authentication.
api_key	Your OmniRouters API key.
wire_api	Set to responses to use the OpenAI Responses API.
supports_websockets	Set to false to disable WebSocket.

3. Restart Codex

Save config.toml, then restart Codex for the configuration to take effect.

If requests show Reconnecting, the configuration usually does not include:

supports_websockets = false

Add the field, save the configuration, and restart Codex again.

Related Links

• AI Apps overview
• CC Switch
• OmniRouters API Reference