Open-source · Codex App & CLI

Run Codex on DeepSeek

Codex Router is a local proxy that adds DeepSeek, Kimi, GLM, Grok and Claude to the Codex App and CLI — native in the model picker, with OAuth or API keys, and your GPT models left untouched.

Codex App & CLI R Codex Router DeepSeek Kimi GLM Grok Claude + more
Live routing map — requests flow to every provider
Codex task
Install the router from this public repository:
https://github.com/duolahypercho/codex-router

Follow AGENTS.md. Preserve my existing Codex models, profiles, settings, and
ChatGPT login. Use only the provider authentication I choose, safely migrate
only recognized older versions, run the Codex doctor, and leave the final app
restart to me. Never ask me to paste a token or API key into chat.
Full quickstart →
Codex App & CLI Router :4102 LiteLLM :4100 DeepSeek · Kimi · GLM · Grok · Claude

Codex speaks the Responses API. The router authenticates the caller, LiteLLM translates the request to each provider’s native protocol, and only the selected provider’s credential leaves your machine. Learn more →

Codex Router is a free, open-source, credential-isolating model router that lets you use DeepSeek, Kimi, GLM, Grok and Claude inside the Codex App and CLI. Instead of replacing your setup, Codex Router merges external models into Codex’s native model catalog, so they appear in the normal model picker right next to your GPT models — with OAuth or API keys, per-provider credential isolation, and your ChatGPT login left untouched. Everything runs locally on 127.0.0.1: the router authenticates the caller, LiteLLM translates each request to the provider’s protocol, and only the selected provider’s credential ever leaves your machine.

Installing Codex Router takes one command — see the install guide for macOS, Linux and Windows — and a doctor command verifies every layer. Each model page (codex deepseek, codex kimi, codex glm, codex grok, codex claude) explains the exact commands, OAuth versus API-key setup, verification steps and common pitfalls. If a model is missing from the picker or a provider stops working, the troubleshooting guide covers the most common fixes, and the changelog tracks every release.

Any provider

DeepSeek, Kimi, GLM, Grok, Claude, opencode Go, Command Code, Meta and 20+ OpenAI-compatible providers.

OAuth or API key

Reuse Kimi or Grok CLI logins, or store keys in hidden per-user prompts — auto-refreshed and isolated per provider.

Native in the model picker

Routed models appear in the Codex App and CLI picker with reasoning levels, next to the native GPT entries.

Your GPT stays

The built-in OpenAI provider, ChatGPT login, profiles and MCP settings are preserved exactly as they were.

Sub-agents

Routed models work as Codex sub-agents, with encrypted task payloads relayed through the native backend.

Local & private

Everything listens on 127.0.0.1. No cloud, no database, no third-party proxy between you and your provider.

Model picker Before
GPT-5.6 Sol OpenAI
GPT-5.6 Terra OpenAI
GPT-5.6 Luna OpenAI
Model picker After
GPT-5.6 Sol OpenAI
GPT-5.6 Terra OpenAI
GPT-5.6 Luna OpenAI
DeepSeek V4 Flash Codex Router
Kimi K3 Codex Router
GLM-5.2 Codex Router
Grok 4.5 Codex Router
Claude Opus 4.8 Codex Router

Native in the model picker

Install the router, restart Codex, and DeepSeek, Kimi, GLM, Grok and Claude appear in your normal model picker — right next to the GPT models you already have. Nothing else in your setup changes.

  • Your GPT models and ChatGPT login stay exactly as they were.
  • OAuth sessions and API keys are isolated per provider.
  • Reasoning-effort levels carry over into the picker.
Browse the model matrix →
DeepSeek Kimi GLM Grok Claude OpenRouter Ollama Cloud Command Code OpenAI + more
Independent community project. Not affiliated with or endorsed by OpenAI, Anthropic, Moonshot AI, DeepSeek, xAI, OpenRouter, opencode or opencodex. Providers may restrict proxy traffic — use at your own risk. About →