Integration · OpenAI

Run Codex CLI through Crazyrouter

Codex CLI supports custom providers in config.toml. The installer adds a Crazyrouter provider, keeps the Responses wire API, and lets you choose a model that supports the Responses API on your upstream channel.

Get an API key GitHub

One-command setup

The recommended path is the current config.toml provider setup, not the older environment-only style. --switch reconfigures an existing install in seconds; --full also installs Node.js and Codex on a clean machine.

1Run the command
Pulls the script from GitHub; nothing else is downloaded.
2Paste your Crazyrouter key
Pick a default model when asked; you can change it later.
3Tool talks to Crazyrouter
Config or env vars are written; the next launch routes through your key.

macOS / Linux

curl -fsSL https://raw.githubusercontent.com/xujfcn/crazyrouter-codex-cli/main/install-crazyrouter-codex.sh | bash -s -- --switch

Windows PowerShell

$env:CRAZYROUTER_CODEX_MODE='switch'; iwr -UseB https://raw.githubusercontent.com/xujfcn/crazyrouter-codex-cli/main/install-crazyrouter-codex.ps1 | iex

What the script does

--switch: Codex is already installed; only the provider config changes
--full: fresh machine — installs Node.js 22 and @openai/codex first
Asks for your key and default model, writes OPENAI_API_KEY and OPENAI_BASE_URL
Adds a [model_providers.crazyrouter] block to ~/.codex/config.toml and sets it as model_provider

Scripts are fetched from the public GitHub repo linked below. Read them before piping to a shell if that's your policy; the download-then-run form is in the README.

Manual configuration

~/.codex/config.toml

toml

model = "gpt-5.6-sol"
model_provider = "crazyrouter"

[model_providers.crazyrouter]
name = "Crazyrouter"
base_url = "https://api.crazyrouter.com/v1"
env_key = "OPENAI_API_KEY"
wire_api = "responses"

wire_api = "responses" keeps Codex on the Responses API, which Crazyrouter relays natively for GPT-5 / Codex models.

Environment variable

bash

export OPENAI_API_KEY=sk-your-crazyrouter-key
# Windows: setx OPENAI_API_KEY "sk-your-crazyrouter-key"

Base URL rule

https://api.crazyrouter.com/v1 — This tool is OpenAI-compatible and appends /chat/completions (or /responses) itself. Use the /v1 base URL exactly as shown.

Recommended models

Live Crazyrouter prices, refreshed hourly. Any model in the catalogue works; these are the ones we'd start with.

GPT-5.6 Sol

gpt-5.6-sol

$2.60 in · $13.00 out /1M tokens

GPT-5.3 Codex

gpt-5.3-codex

$1.40 in · $11.20 out /1M tokens

GPT-5.5

gpt-5.5

$3.25 in · $19.50 out /1M tokens

FAQ

Which wire API should I use?

responses. GPT-5 and Codex models are relayed on the Responses API with reasoning items preserved across turns.

Can I use Claude or DeepSeek from Codex?

Chat Completions support alone does not establish Codex compatibility. Codex requires the Responses API; verify the model and upstream channel before selecting a non-OpenAI model.

Do I still need ChatGPT login?

No. With a custom provider and OPENAI_API_KEY set, Codex never uses the ChatGPT auth flow.

The script says my key is invalid

Make sure the token's model allowlist includes the model you chose and that OPENAI_BASE_URL has no UTM parameters or trailing path beyond /v1.

Guides and comparisons

Podcast

AI Dev Tools: The Crazyrouter Podcast

Short episodes for developers on AI APIs, gateways, cost control and coding tools.

Latest episodes

EP196 · Oct 29, 2026 · 6:53
AI API Security Posture — Turn Controls Into Continuous Evidence
A practical guide to AI API security posture: inventory exposure, verify controls continuously, protect keys and data, constrain tools, and turn security assumptions into evidence that operators can act on.
EP195 · Oct 28, 2026 · 7:21
AI API Policy Drift — Detect When Production Rules Stop Matching Intent
A practical guide to policy drift in AI APIs: version routing and safety rules, detect stale enforcement, compare intended and observed behavior, and restore control before small gaps become incidents.

Episodes with written show notes →