opencodex (lidge-jun/opencodex)

codex-cli
Searching Database…

17.7K ·GitHub
What is this?opencodex is a universal provider proxy that lets OpenAI Codex (CLI, App, SDK) and Claude Code run any LLM — Claude, Gemini, Grok, DeepSeek, Ollama, and 40+ other providers — instead of being locked to one vendor's backend.
Is this a CLI tool or a background service?It's a local proxy run as a background service via the `ocx` CLI, with a web dashboard for configuration — it translates Codex's Responses API into whatever protocol your chosen provider speaks.
How do I get it running?Install it and run `ocx service` to start it in the background, then configure providers and models from the web dashboard (`ocx gui` reopens it anytime); a digest-pinned Docker Compose build is also provided for self-hosting.
Do I need Node or Bun installed first?You need Node 18+; the Bun runtime it relies on internally is bundled automatically by npm install, so there's no separate Bun install and no WSL needed on Windows.
Could routing through this get my Claude account suspended?Possibly — the README explicitly warns that some providers, notably Anthropic, may suspend or restrict accounts detected routing API traffic through third-party proxies, and use is at your own risk.
Which operating systems does it actually run on?macOS (arm64/x64), Linux (x64/arm64), and Windows (x64) are all listed as fully supported, each with their own background service manager (launchd, systemd, or Task Scheduler/WinSW).
What is a 'combo' in opencodex?A combo is one virtual model id that spreads requests across multiple providers using failover or weighted round-robin routing.
What does the ChatGPT account pool feature do?It lets you add several ChatGPT/Codex accounts and auto-route new sessions to the lowest-usage healthy one, while existing threads stay pinned to whichever account started them.
Can sub-agents run on different models than the main agent?Yes — Codex's sub-agent picker can route each sub-agent to its own configured model, with fallback chains if that model fails.
Do I need to paste an API key for every provider?Not necessarily — OAuth login works for xAI, Anthropic, and Kimi, or you can forward 'codex login', paste a key, or use an ${ENVVAR} reference instead.
Creator
lidge-jun
Size
293.6 MB
Stars
17.7K
Age
4mo
Commits · 90 Days