Codex
Add the gateway as a model provider in Codex.
Codex uses the Responses format, so it runs on
/v1/responses and the ChatGPT (Codex) pool.
Add the provider
Add the gateway to ~/.codex/config.toml:
model_provider = "indicia"
model = "gpt-6-luna"
[model_providers.indicia]
name = "Indicia"
base_url = "https://code.test.indicia-ai.com/v1"
env_key = "INDICIA_TOKEN"
wire_api = "responses"env_key names the environment variable that holds your token. Codex sends
it as Authorization: Bearer.
Export the token
export INDICIA_TOKEN="ic_..."Run Codex
codex
codex --model gpt-6-astraChatGPT model ids only
/v1/responses passes the model on unchanged, so use a ChatGPT model id
such as gpt-6-astra or gpt-6-luna. Model classes and Claude models do
not work here.
To try the provider without changing config.toml, pass the same settings
with -c:
codex \
-c 'model_providers.indicia={name="Indicia", base_url="https://code.test.indicia-ai.com/v1", env_key="INDICIA_TOKEN", wire_api="responses"}' \
-c model_provider="indicia" \
-c model="gpt-6-luna"