Use Codex

Overview

This guide explains how to install and configure OpenAI Codex so it can call AI models via the Routin API platform. Codex is a powerful AI coding assistant for multiple languages and environments.

Install Codex

npm install -g @openai/codex@latest

Configure Codex

Step 1: Open the config file

Config locations by OS:

Windows

C:\Users\<your user>\.codex\config.toml

macOS / Linux

~/.codex/config.toml

Tip: If the file is missing, install the VS Code Codex extension and open settings through the extension UI.

Step 2: Edit the config

Open config.toml and add:

model = "gpt-5.2"
model_provider = "meteor-api"
disable_response_storage = true
approval_policy = "never"                  # options: "untrusted" | "on-failure" | "on-request" | "never"
sandbox_mode = "danger-full-access"

[model_providers.meteor-api]
name = "Routin API"
base_url = "https://api.routin.ai/v1"
env_key = "OPENAI_API_KEY"
wire_api = "responses"

Notes:

model: model name to use
model_provider: custom provider name
disable_response_storage: disable storing responses
approval_policy: automation level; never runs without prompts
sandbox_mode: danger-full-access allows full access

Important: model and model_provider must be at the top of the file.

Advanced example: reasoning summaries and higher reasoning effort

model = "gpt-5.2"
model_provider = "meteor-ai"
disable_response_storage = true
approval_policy = "never"                  # options: "untrusted" | "on-failure" | "on-request" | "never"
sandbox_mode = "danger-full-access"

rmcp_client = true
model_reasoning_effort = "xhigh"

# Reasoning summary: auto | concise | detailed | none (default: auto)
model_reasoning_summary = "detailed"

# Text verbosity for GPT-5 family (Responses API): low | medium | high (default: medium)
model_verbosity = "high"

# Force-enable reasoning summaries for the current model (default: false)
model_supports_reasoning_summaries = true

[mcp_servers.claude]
command = "claude"
# Optional
args = ["mcp", "serve"]

[model_providers.meteor-ai]
name = "meteor-ai"
base_url = "https://api.routin.ai/v1"
env_key = "METEOR_AI_API_KEY"
wire_api = "responses"

Step 3: Configure the API key

Store the API key as a system environment variable (avoid local files).

Windows (PowerShell)

setx OPENAI_API_KEY "sk-your-api-key-here"

macOS / Linux

export OPENAI_API_KEY="sk-your-api-key-here"

Replace sk-your-api-key-here with your Routin API key. Restart the terminal after setting it, or add the command to ~/.zshrc / ~/.bashrc for persistence.

Verify installation

After configuring, open a terminal:

codex

If everything is correct, Codex starts in interactive mode and is ready to use.

FAQ

Config file missing

If you cannot find the config file, create it manually:

Windows

# Create config directory
mkdir %USERPROFILE%\.codex

# Create config file
notepad %USERPROFILE%\.codex\config.toml

macOS / Linux

# Create config directory
mkdir -p ~/.codex

# Create config file
nano ~/.codex/config.toml

How do I get an API key?

See the “Get API Key” section in the Claude Code guide.

Get Help

Web: https://routin.ai

Use Codex

On this page