Codex Router

Codex Router runs on your computer and connects the Codex App and CLI to external model providers or your own self-hosted endpoints. It's for developers who want a different model within their existing coding workflow. Cloud routes send model requests to remote services and need internet access; paid providers still charge for access or usage.

Supported providers include Anthropic's Claude, Kimi, DeepSeek and xAI's Grok. Ollama Cloud routes use an ollama.com account and run inference remotely. Custom endpoints can point to a server you control and support Chat Completions or Responses, so where inference happens depends on the endpoint you choose.

Provider credentials stay in local storage, though the router uses them to authenticate requests to the selected service. Kimi and Grok can use existing official CLI OAuth sessions or separate API credentials. Kimi's global and mainland China API services have distinct accounts and keys.

The local Control Center manages providers, endpoint catalogs and which models appear in client pickers. The catalog accounts for available credentials, while GitHub Copilot routes select account-visible models with Responses API, streaming and tool-call support. Native GPT models continue through Codex directly.

The same router can serve DeepSeek Harness, Gemini CLI, Cursor Agent, Cursor App, Claude Code and OpenClaw. It's open source under the MIT license and runs on macOS, Windows and Linux; Linux support targets the Codex CLI. GitHub Copilot support targets GitHub.com and excludes Enterprise Cloud data-residency hosts.