Overview
codex-keysmith installs a standing instruction onto a local Codex setup and, just as importantly, takes it back off again. Its whole premise is that changing how your coding agent behaves should be a reviewable operation: every command prints the write plan first and touches nothing until you confirm, and removal follows the same preview-then-confirm path back to the original state.
It does not patch Codex itself and it never reads accounts or API keys — it writes into the Codex configuration directory you point it at, and the effect applies to conversations you open afterwards, not the one you are in. A default install ships a single prompt. The project describes the surrounding machinery as versioned deployment with ownership manifests, hook isolation, scenario evaluation and recovery, which is what separates it from copying a prompt file into place by hand.
Distribution is deliberately conservative: a single-file Python script attached to each stable GitHub Release, published alongside a SHA256SUMS file so you can verify the download before running it. Unsigned desktop builds exist for Apple Silicon and Windows x64. Keysmith is a family — sibling installers cover Claude Code, Grok Build and ZCode — with one installer per tool. It needs Python 3.10+ and is MIT-licensed.
What it does
- Dry-run first: every command shows the target directory, the prompt source and the write plan before anything is written
- Reversible by design — review the removal plan, confirm, and the machine returns to its previous state
- Leaves the Codex app untouched and never reads accounts or keys; only the configuration directory you name is written
- Versioned deployment with ownership manifests, hook isolation, scenario evaluation and recovery
- Shipped as a single-file script per release with a SHA256SUMS companion for integrity checking
- One installer per agent: sibling projects cover Claude Code (claude-keysmith), Grok Build and ZCode
Getting started
Codex must already be installed. The recommended path is the single-file script from the latest stable release; download it with its checksum file and verify before running.
Download and verify the release script
Replace vX.Y.Z with the newest stable tag on the Releases page.
base='https://github.com/Jia-Ethan/codex-keysmith/releases/download/vX.Y.Z'
curl --fail --location --remote-name "$base/codex-instruct-vX.Y.Z.py"
curl --fail --location --remote-name "$base/SHA256SUMS"
awk '$2 == "codex-instruct-vX.Y.Z.py"' SHA256SUMS | shasum -a 256 -c -Check the current state
Point it at your Codex configuration directory and ask what is installed today.
python3 codex-instruct-vX.Y.Z.py --version
python3 codex-instruct-vX.Y.Z.py --codex-dir ~/.codex --statusPreview the write plan
Nothing is written in this mode. Confirm the target directory, the prompt source and the planned changes before applying.
python3 codex-instruct-vX.Y.Z.py --codex-dir ~/.codex --dry-runStart a new conversation
The instruction applies to conversations opened after the install, so close the old task and open a fresh one.
Commands and code are distilled from the project's own documentation — always check the official repo for the latest.
When to use it
- Roll a shared team instruction out to Codex on many machines and be able to prove what each write changed
- Try a new standing prompt knowing the removal path is the same reviewed, reversible operation
- Keep agent instructions versioned and verifiable instead of hand-edited config files that drift
- Audit what a prompt install would touch before allowing it on a work machine
How codex-keysmith compares
codex-keysmith alongside other open-source coding-agent harnesses & runtimes tools AI/TLDR tracks, ranked by GitHub stars.
| Tool | Stars | What it does |
|---|---|---|
| CC Switch | ★ 133k | A cross-platform desktop app that manages provider configuration for eight coding-agent CLIs from one place, with a local failover proxy, MCP and skill management, and spend tracking. |
| Oh My OpenAgent | ★ 69.2k | Opinionated agent bundle for OpenCode and Codex CLI that ships preconfigured sub-agents, lifecycle hooks and built-in MCP servers behind a single ultrawork command. |
| herdr | ★ 39.3k | A single-binary terminal multiplexer built for coding agents: persistent sessions across local and SSH machines, per-pane working/blocked/idle status, and a socket API agents drive themselves. |
| oh-my-claudecode | ★ 39.2k | A plugin and CLI that adds multi-agent orchestration to Claude Code: staged autopilot workflows, parallel tmux workers, and advisors that route a second opinion to Codex, Gemini, Grok or Cursor. |
| AionUi | ★ 32.9k | An open-source desktop workspace that pairs a built-in agent engine with dozens of external CLI agents, adding scheduled runs, MCP tools and remote access from a WebUI or chat apps. |
| cmux | ★ 27.2k | A Ghostty-based macOS terminal built around coding agents: vertical tabs that show git branch, PR status and ports, attention notifications when an agent needs you, and a scriptable built-in browser. |
| cc-connect | ★ 15.5k | A Go daemon that bridges local coding agents — Claude Code, Codex, Cursor Agent, Gemini CLI, OpenCode, Kimi CLI and more — to Feishu, DingTalk, Slack, Telegram, Discord, LINE and WeCom, so you drive them from chat. |
| codex-keysmith | ★ 4.5k | Install a reversible instruction onto Codex — preview, apply, undo |