Codex CLI

Codex CLI uses OpenAI's Responses API, which we serve at https://pinerouter.ai/v1/responses.

1. Install

npm install -g @openai/codex

brew install --cask codex

2. Point it at us

Add a provider to ~/.codex/config.toml:

model = "gpt-5-5"
model_provider = "ai"

[model_providers.ai]
name = "AI"
base_url = "https://pinerouter.ai/v1"
env_key = "AI_API_KEY"
wire_api = "responses"

Then put your key in the environment named by env_key:

echo 'export AI_API_KEY="sk-..."' >> ~/.zshrc   # or ~/.bashrc
source ~/.zshrc

setx AI_API_KEY "sk-..."

3. Run

cd your-project
codex

Switch models inside Codex with /model, using any GPT model id from Models.

Add our media models as tools

[mcp_servers.ai]
url = "https://pinerouter.ai/mcp"
http_headers = { Authorization = "Bearer sk-..." }

See MCP.

Troubleshooting

• 401: check that AI_API_KEY is set in the terminal you run codex from.
• Model not found: use a model id exactly as listed on Models.
• Stream errors: make sure wire_api = "responses" and base_url ends in /v1.