Running the Codex CLI on Claude

Codex is OpenAI's agentic coding CLI, but its provider system is open — define a custom endpoint in config.toml and it drives Claude through AI Prime Tech instead of OpenAI. One gateway, every model (Claude, GPT, Gemini) behind a single base URL.

Add a custom provider

Codex reads ~/.codex/config.toml, shared by the CLI, the SDK, and the VS Code extension. Add a provider table and point Codex at it.

1. Open ~/.codex/config.toml (create it if missing).
   terminalCopy

   $EDITOR ~/.codex/config.toml

2. Add a [model_providers.<name>] table with the gateway base_url, the env var holding your key, and wire_api = "chat" (the OpenAI Chat Completions shape the gateway accepts for Claude).
   ~/.codex/config.tomlCopy

   model = "claude-sonnet-4-5"
   model_provider = "aiprime"
   
   [model_providers.aiprime]
   name = "AI Prime Gateway"
   base_url = "https://aiprimetech.io/v1"
   env_key = "AIPRIME_API_KEY"
   wire_api = "chat"

3. Export your key into the variable named by env_key, then launch Codex.
   terminalCopy

   export AIPRIME_API_KEY=<your key>
   codex

Pick a model

The model field is the plain id the gateway serves — no provider prefix. Because AI Prime Tech is multi-model, you switch between a Claude, GPT, or Gemini id by changing one value.

Why route Codex through a gateway

Codex tasks are multi-turn loops — read, edit, run, re-read — and each turn resends context, so one task moves large token volume. Routing through AI Prime Tech gives a single billing surface and clean per-session attribution across your whole team, plus the freedom to swap models without juggling separate provider keys.

Frequently asked questions

AI Prime Tech is an independent API gateway. It is not affiliated with, endorsed by, or a reseller of Anthropic. Claude and related model names are trademarks of their respective owners.