Codez

A Codex-compatible runtime fork for local agents, token-aware context control, App Server v2, and plugin hooks.

Codez keeps upstream Codex compatibility where it matters, then layers focused runtime experiments on top.

Compatibility · Stack · RTK Plugin · Pitlane Plugin · Getting Started · Config

Codez is a fork of OpenAI Codex. The goal is not to hide that lineage: upstream Codex remains the base CLI, protocol, package shape, and user-facing mental model. Codez is the fork layer for local-agent runtime work that benefits from moving faster than upstream.

Why Codez Exists

• keep a Codex-compatible CLI/runtime while experimenting with local-agent needs
• reduce wasted tokens from stale tool, reasoning, and context history
• support plugin and hook workflows that can harden shell usage
• expose App Server v2 surfaces for local clients and future gateways
• make context-pressure and prompt-history behavior easier to evolve
• provide a clear core layer for future gateway projects without baking a specific gateway into the runtime

Core Additions

Codez is not only a renamed README around upstream Codex. The fork carries runtime work aimed at long local-agent sessions:

• prompt-history pruning before model sampling: stale context messages, older reasoning items, tool calls, and tool outputs can be removed before the next request while preserving the active turn and image-output dependencies
• goal-continuation pruning: automatic goal follow-up turns are treated as fresh prompt boundaries, so previous tool-heavy work can be trimmed while the active goal objective remains visible
• live-steer prompt boundary safety: tiny steering updates during a run can stay adjacent to the active turn without disabling broader stale-history pruning for later requests
• compaction-aware pruning: the remote/autocompact path can apply the same pruning and trim function-call history before compaction, reducing context pressure before a summarization turn
• App Server v2: a local app/server protocol surface for richer clients, gateway experiments, thread operations, goal RPC, hook/catalog inspection, and command event flows
• plugin hook compatibility: Codez keeps plugin-loaded hook paths usable for PreToolUse workflows, so optional plugins such as RTK and Pitlane can run as normal runtime extensions

Part of the Codez stack

The Codez stack is modular. Each layer can be used on its own unless a higher layer explicitly opts into it.

Layer	Public surface	Responsibility	Dependency
Codez	Codex-compatible runtime	App Server v2, goal RPC, long-session hardening, prompt pruning, and plugin hooks	Does not require Teledex
RTK Codex Plugin	Optional Codex plugin	Shell/token safety through rtk rewrite and bounded output guarding	Requires a Codex-compatible plugin-hook runtime; does not require Teledex
Pitlane Codex Plugin	Optional Codex plugin	Code-navigation/token-saving rewrites through a host-local pitlane CLI	Requires a Codex-compatible plugin-hook runtime and local pitlane; does not require Teledex
Teledex (planned public repo: Krablante/teledex)	Telegram gateway/session layer	Topics, queues, live steer, /goal UX, and multi-host delivery/recovery	Basic mode can drive upstream Codex; full mode requires a Codez-compatible runtime with App Server v2 and plugin-hook support

What Is Different

Codez keeps upstream names such as codex, @openai/codex, and Codex protocol terms where compatibility matters. Fork-specific docs use the Codez name when describing additions, release projections, or stack positioning.

RTK and Pitlane are published separately:

https://github.com/Krablante/rtk-codex-plugin https://github.com/Krablante/pitlane-codex-plugin

The Telegram gateway layer is not linked here yet because its public repo has not been published. The planned public name is Krablante/teledex.

Quick Start

Packaged Codez releases are a publication decision. Until then, use the source tree directly:

git clone https://github.com/Krablante/codez
cd codez
pnpm install
just install
just codex --help

For upstream Codex usage, package names and official setup are still documented by OpenAI:

• Codex documentation
• Upstream repository

Read Next

• Compatibility matrix
• Codez stack
• Optional RTK plugin integration
• Optional Pitlane plugin integration
• Installing and building
• Configuration
• Contributing

License and Attribution

Codez is distributed under the same Apache-2.0 license as upstream Codex. OpenAI Codex remains the upstream project; Codez is an independent fork projection and is not presented as an official OpenAI release.