Codex CLI integration

Codex CLI supports custom providers through ~/.codex/config.toml. The two most common integration methods are shown below.

One-click import (recommended)

The DimiLinks console can send a GPT key directly to CC Switch:

On the API Keys page, click “Import to CC Switch” for a key with no model restriction or a restriction to GPT models.
CC Switch automatically adds a Codex provider with the base URL https://api-direct.dimilinks.com/v1.
Before starting Codex, run cc-switch use codex to route Codex through the DimiLinks OpenAI-compatible endpoint.

To manage it yourself, edit ~/.codex/config.toml and add a provider:

[providers.dimilinks]
name = "DimiLinks"
base_url = "https://api-direct.dimilinks.com/v1"
api_key = "sk-..."
default_model = "gpt-5.5"

Then switch to it in Codex:

codex provider use dimilinks
codex

To use DimiLinks by default, set provider = "dimilinks" at the top level of the configuration.

export OPENAI_BASE_URL="https://api-direct.dimilinks.com/v1"
export OPENAI_API_KEY="sk-..."
codex

This is useful for temporary troubleshooting or CI.

Treat GET https://api-direct.dimilinks.com/v1/models as the authoritative model catalog.

401 Invalid token: Check whether shell quoting truncated api_key / OPENAI_API_KEY.
404: This occurs when the base URL is set to https://api-direct.dimilinks.com/ without /v1. The OpenAI-compatible endpoint requires /v1.
429 rate_limit_rpm: Reduce concurrency. OPENAI_MAX_RETRIES already backs off by default, but you can increase it.
Keep an official OpenAI account available as well: Keep another provider in config.toml and switch with provider use when needed.