# Codex CLI

> Run OpenAI's Codex CLI and the Codex extension against the gateway.

_Source: https://aiprimetech.io/docs/guides/codex-cli/ · Home > Docs > Integration guides_

Codex CLI reads the standard OpenAI environment variables, so redirecting it needs no patch.

```bash
export OPENAI_BASE_URL="https://aiprimetech.io/v1"
export OPENAI_API_KEY="sk-your-key"

codex
```


## Config file

```bash
# ~/.codex/config.toml
model = "gpt-5.5"
model_provider = "claudeapikey"

[model_providers.claudeapikey]
name = "ClaudeAPIKey.dev"
base_url = "https://aiprimetech.io/v1"
env_key = "OPENAI_API_KEY"
```

Both GPT and Claude ids work here — the gateway translates, so you can run Codex CLI against `claude-sonnet-4-6` if you prefer it.

> Codex is an agent and streams heavily. If output arrives in one lump rather than progressively, a proxy between you and the gateway is buffering — see [Streaming](/docs/api-reference/streaming/).

- [Chat Completions](https://aiprimetech.io/docs/api-reference/chat-completions/) — Endpoint reference
- [Models](https://aiprimetech.io/docs/getting-started/models/) — GPT and Claude ids


---
_ClaudeAPIKey.dev is an independently operated, Anthropic-compatible API gateway. Not affiliated with Anthropic._