Codex

Connect the Codex CLI (OpenAI’s codex) to Routero by defining a custom model provider in its config. Codex then sends every request through your gateway.

Codex supports two wire formats — chat (/chat/completions) and responses (/responses). Routero supports both; responses is recommended for Codex.

What you must set

In ~/.codex/config.toml:

model_provider = "routero"
model = "openai/gpt-5.5"
preferred_auth_method = "apikey"

[model_providers.routero]
name = "routero"
base_url = "https://api.routero.dcsmartvision.com/v1"
wire_api = "responses"

Then provide your Routero key as the API key:

export OPENAI_API_KEY="YOUR_ROUTERO_KEY"
Setting Value / meaning
model_provider the name of the [model_providers.*] block below (routero)
model any model Routero serves (e.g. openai/gpt-5.5)
preferred_auth_method apikey — authenticate with an API key, not ChatGPT login
base_url https://api.routero.dcsmartvision.com/v1
wire_api responses (recommended) or chat
OPENAI_API_KEY your Routero virtual key (sent as Authorization: Bearer)

preferred_auth_method = "apikey" tells Codex to authenticate with OPENAI_API_KEY instead of ChatGPT sign-in. Set that env var to your Routero virtual key, not an OpenAI key.

Create a key for each developer

In the dashboard, open API Keys and create a virtual key per developer — scoped to their team, limited to the approved models, with an optional budget.

→ Calling the API for the base URL and auth model. → Claude Code and Cursor for the other agents.