OpenLuxRun the Codex CLI on OpenLux models
The Codex CLI speaks any OpenAI-compatible endpoint. One provider block in ~/.codex/config.toml lets your terminal agent run on GPT-6 Astra, Claude Sonnet 4.5, DeepSeek V4 or any other OpenLux model — metered per token on one key.
Set it up in four steps
Install the Codex CLI
Install with npm (npm install -g @openai/codex) or Homebrew, then run it once so the config directory is created.
Create an OpenLux key
Sign in to the console and create an API key. Billing is pay as you go, and free trial credit covers your first runs.
Register OpenLux as a provider
Open ~/.codex/config.toml and add a model_providers entry with the OpenLux base URL and an environment variable for the key.
Run Codex on OpenLux
Export OPENLUX_API_KEY and start codex. Switch models with the --model flag or by editing one line of the config; everything else works exactly as before.
Configuration
Copy these blocks, replace the placeholder key and you are done.
~/.codex/config.toml
model = "gpt-6-astra"
model_provider = "openlux"
[model_providers.openlux]
name = "OpenLux"
base_url = "https://openlux.lol/v1"
env_key = "OPENLUX_API_KEY"
wire_api = "chat"Terminal
export OPENLUX_API_KEY="sk-openlux-..."
codexModels that fit OpenAI Codex CLI
GPT-6 Astra
OpenAI · $0.3677 / 1M tokens
Claude Sonnet 4.5
Anthropic · $0.2647 / 1M tokens
DeepSeek V4
DeepSeek · $0.0613 / 1M tokens
Gemini 3.8 Flash
Google · $0.0551 / 1M tokens
The full list is in the model catalogue and the use-case collections.
OpenAI Codex CLI + OpenLux questions
Does the Codex CLI work with any OpenAI-compatible gateway?
Yes. Codex resolves model providers from ~/.codex/config.toml, and a provider block with a base_url and env_key is all it needs. OpenLux serves /v1/chat/completions, so wire_api stays "chat".
Which OpenLux models are best for terminal coding agents?
Claude Sonnet 4.5 and GPT-6 Astra lead for code edits and planning, DeepSeek V4 and Qwen3 Max are strong budget picks, and the flash-tier Gemini models keep long agent loops cheap.
How do I switch models mid-project?
Either edit the model line in config.toml or pass --model at launch, for example codex --model claude-sonnet-4-5-20250929. The provider stays the same, so context and workspace settings are unaffected.
What does running Codex through OpenLux cost?
Per-token rates from $0.0368 per 1M tokens (Gemini 3 Flash preview) up to $0.4412 (Claude Opus 5), all discounted against official pricing, with every request itemised in the console.
How much will this cost?
Estimate Codex CLI cost
Model tasks, tokens, retries and monthly team spend before rollout.
Open the cost guideAdd AI review to the pull-request loop
Build a bounded code reviewer
Use the same OpenAI-compatible endpoint to review pull-request diffs, publish findings and estimate the review budget.
Open the AI code review guideConnect OpenAI Codex CLI to OpenLux today
One key covers every model in this guide and the rest of the catalogue, metered in USD with free trial credit to start.