Codex CLI

Codex is OpenAI's official terminal coding assistant. It lets you define custom model providers in its configuration file — just point one at FIM Gate.

Setup

1. Install Codex:

   # Homebrew
   brew install codex
   
   # or npm
   npm install -g @openai/codex

2. Edit the configuration file ~/.codex/config.toml (create it if it doesn't exist):

   # Default model
   model = "gpt-5.2"
   # Default model provider; must match a key under [model_providers.*] below
   model_provider = "fim-gate"
   
   [model_providers.fim-gate]
   # Provider name shown in the Codex UI
   name = "FIM Gate"
   # Endpoint Codex will send requests to
   base_url = "https://api.gate.fim.ai/v1"
   # Name of the environment variable holding your API key; its value is sent as a Bearer token
   env_key = "FIM_API_KEY"
   # Protocol type, either "chat" or "responses"; defaults to "chat" if omitted
   wire_api = "chat"

3. Set the environment variable (the name must match env_key and the value must not be empty):

   export FIM_API_KEY="key created in the Console at https://api.gate.fim.ai/dashboard"

   Add it to ~/.zshrc or ~/.bashrc to make it persistent.

4. Start using it:

   # Interactive session
   codex
   
   # One-off question
   codex "Show me how to run the tests in this repo"

You can set model to any available model; for models based on the Responses API, change wire_api to "responses".