Codex++

中文 | English

Codex++ is an external enhancement launcher and manager for the Codex App. It does not modify the original Codex installation. Instead, it starts Codex externally and injects enhancements through the Chromium DevTools Protocol.

Quick Start

Download the latest installer from GitHub Releases:

• Windows: CodexPlusPlus-*-windows-x64-setup.exe
• macOS Intel: CodexPlusPlus-*-macos-x64.dmg
• macOS Apple Silicon: CodexPlusPlus-*-macos-arm64.dmg

After installation, two entry points are available:

• Codex++: a silent launcher. It does not show the manager UI and only starts Codex with Codex++ injection.
• Codex++ Manager: a Tauri control panel for launch, diagnostics, repair, updates, relay injection, enhancements, and user scripts.

The Windows installer creates desktop and Start Menu shortcuts. The macOS DMG installs /Applications/Codex++.app and /Applications/Codex++ 管理工具.app.

Sponsors

JOJO Code | Official Codex++ Relay

The official Codex++ relay service, focused on stable access and cost-effective pricing. JOJO Code supports GPT-5.5, GPT-5.4, Claude Opus 4.8, Claude Opus 4.7, gpt-image-2, and more for daily development, team collaboration, and long-running project workflows.

Want to be shown below?

🏆 Sponsor 🏆	Introduction
	JOJO Code | Official Codex++ Relay Thanks to JOJO Code for sponsoring this project. JOJO Code is the official Codex++ relay service with cost-effective pricing and stable, easy-to-configure Codex API access. It supports GPT-5.5, GPT-5.4, Claude Opus 4.8, Claude Opus 4.7, gpt-image-2, and more for daily development, quick setup, team collaboration, and continuous use.
	AIGoCode Thanks to AIGoCode for sponsoring this project! AIGoCode is an all-in-one platform integrating the latest Claude Code, Codex, and Gemini models, providing stable, efficient, and cost-effective AI programming services. It offers flexible subscription plans, direct access in China, no extra network setup, and fast responses. AIGoCode provides a special benefit for CodexPlusPlus users: users who register through this link can receive an extra 10% bonus credit on their first recharge.
	APIKEY.FUN Thanks to APIKEY.FUN for sponsoring this project! APIKEY.FUN is an AI relay platform focused on open, stable, and cost-effective access to mainstream global models. It supports API relay services for Claude, OpenAI, Gemini, and other popular models, with prices as low as 7% of the official rate. Register through the dedicated link to receive up to a permanent 5% recharge discount.
	RunAPI Thanks to RunAPI for sponsoring this project! RunAPI is an efficient and stable OpenRouter alternative API platform. One API key can access OpenAI, Claude, Gemini, DeepSeek, Grok, and 150+ mainstream models at prices as low as 10% of the original rate, with seamless compatibility for tools such as Claude Code and OpenClaw.
	0029 Cloud Bridge | Codex API Relay Station (gpt5.5 gpt-image-2) Supports individual and enterprise access. Monthly plans and pay-as-you-go billing are available, with Pro/Plus account pools, stable site-wide APIs, and 24/7 technical support.
	Codelf Codelf is a desktop app with an autonomous AI Agent and a full editor. It can help users build projects, organize materials, operate local apps, and work with multiple AI model providers through natural language, with direct access in China and high context-cache efficiency.
	Yiyun Technology Yiyun Technology provides payment and settlement products for AI aggregation businesses, including Jiuwu Yunshang and Yiyun Pay. It supports WeChat Pay, Alipay, UnionPay, and Cloud QuickPass channels with low rates, D1/D0 settlement, 24/7 technical support, dedicated WeCom service groups, and strong website protection for merchants.
	Sui Xiang AI Gateway Thanks to Sui Xiang AI Gateway for sponsoring this project! Sui Xiang AI Gateway is a reliable and efficient API relay service provider for Claude, Codex, Gemini, and more. It focuses on privacy, transparent service, fast support, no data resale, and no model dilution. New accounts can receive 0.5 CNY in daily check-in test credit, with 1:1 recharge credit, no subscription, and pay-as-you-go billing.
	BytePlus ModelArk | Dola Seed Thanks to Dola Seed for sponsoring this project! Dola Seed 2.0 is a full-modal general large model independently developed by ByteDance for the global market. Built on a unified multimodal architecture, it supports joint understanding and generation of text, images, audio, and video. It natively enables agent collaboration, strong reasoning, long-task execution, tool integration, and coding capabilities, and is readily accessible through the ModelArk platform. Register via this link to get 500,000 tokens of free inference quota per model. Mainland China developers can click here.
	Smallice | AI Relay Thanks to Smallice for sponsoring this project! Smallice is one key to all the language models worth calling: a unified endpoint that acts as a quiet foundation layer beneath your applications. Whether you call Claude, GPT, Gemini, or DeepSeek, the request shape stays consistent. Register through this link to get started.

Highlights

• Rust backend and silent launcher with no extra runtime requirement.
• Tauri + React manager with dark/light theme support.
• External CDP injection. No app.asar patching and no DLL writes into the Codex installation.
• Relay injection mode with multiple relay profiles, CodexPlusPlus provider configuration, and a one-click switch back to official ChatGPT login mode.
• Traditional enhancement mode with plugin marketplace unlock, session delete, Markdown export, project move, and more.
• Paste fix: when pasting from Word or other rich-text sources into the Codex composer, only keep the plain text so Codex does not treat the clipboard content as an image or file attachment. Off by default; requires a Codex relaunch to take effect.
• Usage note: after toggling in the manager, click the "保存增强设置" / "Save enhancement settings" button to persist, then restart Codex++ for the change to take effect.
• Independent user script management with startup injection.
• Provider Sync to keep historical sessions visible after switching providers.
• Zed open entry detects remote SSH context and opens the matching remote file in Zed Remote Development from Codex.
• Per-model context window configuration: the "Model list" is split into two columns, model name on the left and context window (e.g. 1M, 200K, or 1000000) on the right. Codex++ auto-generates model_catalog_json and injects it into config.toml; the matching window is applied when you switch models. Leave the window empty to use Codex's default length.
• Upstream worktree creation: create new worktrees from upstream/<base-branch> after fetching the remote branch, reducing conflicts caused by stale local HEAD state.
• GitHub Release updates. Both the manager and silent launcher can detect available updates.
• Windows single instance, no console window, administrator manifest, and system Desktop path detection.
• Separate macOS x64 and arm64 DMGs. The silent launcher hides its Dock icon.

Relay Injection

Relay injection is for users who are already logged in with an official ChatGPT account in Codex/ChatGPT and want model requests to go through a custom compatible API.

The boundary of this hybrid mode is:

• The official ChatGPT/Codex login state still owns Codex App account features and the plugin entry.
• The relay profile only controls the Base URL, key, and model names used for model requests.
• The compatible API provider is not tied to any specific vendor; it only needs to match the selected upstream protocol and Codex configuration.
• Clearing API mode should return Codex to the official login mode so the official account and plugins keep working.

Before applying relay injection, run a minimal preflight:

1. Make sure Codex has detected the ChatGPT login state and the plugin entry is available.
2. Confirm the custom Base URL is reachable and supports the selected upstream protocol, such as a Responses-compatible endpoint.
3. Test the target key with the smallest useful auth probe, such as a model-list request or a short message request.
4. Only record whether the key exists and whether auth passed. Do not paste real keys into logs, screenshots, or issues.
5. Make sure ~/.codex/config.toml has a backup so clearing API mode can safely roll back.

In the manager's Relay Injection page:

1. Make sure ChatGPT login status is detected.
2. Add one or more relay profiles with Base URL and Key.
3. Select the active profile and apply relay injection.
4. Launch Codex++.

Codex++ writes configuration similar to this into ~/.codex/config.toml:

model_provider = "CodexPlusPlus"

[model_providers.CodexPlusPlus]
name = "CodexPlusPlus"
wire_api = "responses"
requires_openai_auth = true
base_url = "https://example.com/v1"
experimental_bearer_token = "sk-..."

To return to the official login mode, use the clear API mode button in the Relay Injection page. This removes OPENAI_API_KEY related configuration and switches Codex back to official ChatGPT authentication.

Enhancements

Enhancements are controlled in the manager. Enhancement injection is enabled by default. When disabled, Codex++ will not inject its menu or scripts.

When relay injection mode is active, plugin marketplace unlock is unnecessary, and the UI will say so. Other enhancements, including session delete, export, move, paste fix, recommendations, and user scripts, can still be used.

Recommendations

Re