CC Simplerouter

simplerouter for Codex CLI

simplerouter launches an installed OpenAI Codex CLI
with an existing Codex subscription or against OpenRouter, Google AI Studio, OpenAI,
xAI, DeepSeek, Z.AI, Meta Model API, or a local LM Studio server.
It supplies Codex with one session-scoped Responses provider and a model descriptor
derived from the installed Codex release, preserving Codex features such as freeform
apply_patch, namespaced tools, parallel tool calls, reasoning, and multi-agent support.

The launcher does not rewrite ~/.codex/config.toml or replace the user's Codex
installation. Provider overrides, temporary model metadata, and any localhost proxy
exist only for the child Codex process and are removed when that process exits.
Launched sessions also use Codex's standard service tier, disabling Fast mode without
changing the user's global service-tier preference for direct Codex sessions.
When the patched Codex companion bundle is installed, launched sessions stream
reasoning in the transcript by default and expose /thinking as a session-only toggle.

Requirements and installation

Install a current Codex CLI first. OpenAI's standalone installers are:

Windows:

macOS/Linux:

Then build and install simplerouter from this repository. This requires Go 1.24
or newer.

Windows:

macOS/Linux:

Both scripts install the binary under ~/.local/bin and explain any required
PATH change.

To prepare, build, and install the patched Codex companion on Windows, run:

The build script clones the pinned upstream Codex release into ignored
.build/, applies the patch series committed under codex/patches, and
verifies the exact resulting Git tree before compiling. Cargo outputs are kept
separately in .build/codex-target, so source refreshes retain incremental
build artifacts. It installs a…