SuperGness/codey

What is Codey?

Codey is a desktop‑side enhancement layer for the Codex client (the GUI that lets you talk to OpenAI‑style models). It adds a control panel where you can:

Switch between different model endpoints (official OpenAI, third‑party services, self‑hosted gateways).
Manage multiple ChatGPT accounts, see quota/refresh times, and even sign‑in via the “Sign in with ChatGPT” flow.
Track usage statistics – tokens, latency, error rates, and model‑wise breakdowns.
Organise conversations, import/export sessions, delete or restore specific turns.
Install, enable, and configure extensions (MCP, Skill, “trusted Codey” plugins) and view their logs.
Receive task notifications through Feishu, WeChat Work, Telegram, ntfy, or a custom WeChat bot.
Run simple desktop‑automation actions (click, type, scroll) via the optional Codey Computer Use plugin.
Optimize prompts with a one‑click “prompt‑enhance” tool.
Use the FastCtx file‑context helper to search and batch‑replace across files.
Perform diagnostics, clean logs, and choose between stable, experimental, or “no‑check” update channels.

In short, Codey turns the Codex GUI into a multi‑account, multi‑model hub with analytics, plug‑in management, and lightweight RPA capabilities.

Who might use it?

User	Why it helps
Developers who rely on Codex for code generation and want to switch models or accounts without restarting the client.
Teams that need to monitor token consumption across projects and get alerts when a job finishes or fails.
Power users who like to extend Codex with custom plug‑ins, automate UI actions, or batch‑edit code bases via FastCtx.
Researchers experimenting with different LLM providers and wanting a single pane to compare performance and cost.

Key Features (as described in the README)

Category	Features
Model & Endpoint Management	Add official and third‑party routes, configure proxies, drag‑reorder models, switch model and “thinking depth” per task.
Remote Compression / Summarisation	Use upstream compression or gateway summarisation; can generate summaries with a new account while preserving full history if capacity permits.
Account Handling	Store multiple ChatGPT accounts, view quota and reset timers, switch accounts mid‑conversation, support OpenAI direct‑auth.
Usage Analytics	Show route selection, retries, latency, output speed, token counts, errors; trend analysis and model‑share breakdowns.
Session Management	New sessions default to medium inference strength; import/export, delete specific turns, restore from backup.
Extension (Plug‑in) Management	Manage MCP, Skill, and “trusted Codey” plug‑ins; configure, enable/disable, drag‑drop install, view/clean logs.
Plugin Routing	Manually enter or rotate keys, bind saved accounts, obtain request status via a secondary account before the official request.
Task Monitoring	Query number of running/failed tasks via an auth API.
UI Enhancements	Improved plug‑in marketplace, quick access to total turns, cumulative tokens, cache hits, average speed of the latest turn.
Desktop Automation	Optional Codey Computer Use plug‑in can read app state and perform clicks, typing, scrolling.
Prompt Optimisation	One‑click optimisation of input prompts, editable result.
Model Collaboration	Sub‑agents can edit different files in parallel; roles, models, and depth can be tweaked live.
File Tools	Optional FastCtx integration for reading, searching, and bulk‑replacing file contents.
Notifications	Send completion/failure/await‑intervention alerts via Feishu, WeChat Work, Telegram, ntfy, or WeChat ClawBot.
Diagnostics & Updates	Config checks, auto‑reconnect, log cleaning, three‑tier update strategy (stable, experimental, no‑check).
Release Management	View device ID, package specific versions, publish AI changelogs, roll back to previous stable releases.

How to get started (from the README)

Launch Codey – it automatically starts the Codex client.
Click the Codey button in the left‑hand icon bar to open the control console.
Configure your accounts and routes (add official accounts, set proxies, etc.).
Optionally enable the Codey Computer Use plug‑in (requires macOS 14+ or Windows, plus accessibility/screen‑recording permissions on macOS).
Use the UI to start a new conversation, pick a model, and watch the usage panel for token/latency stats.
Install extra plug‑ins from the built‑in marketplace if you need file‑context tools or other extensions.

Limitations & Caveats (as noted by the authors)

Works only with the official Codex desktop client – it will restart the client when changing settings.
Each route uses its own account; the “default” account determines the login identity for the client.
Local routing can set a context budget but cannot increase the provider’s quota.
Cross‑route compatibility depends on the provider and Codex version; full history may not always be recoverable.
Native plug‑ins run with the same system permissions as Codey, so only install from trusted sources.
macOS requires explicit accessibility and screen‑recording permissions; upgrades may need re‑authorisation.
“No‑check” disables automatic update checks but you can still manually check stable releases.
Experimental builds may have data‑compatibility issues – back up important work before enabling.

License & Credits

Codey itself is maintained by SuperGness (contact: kimzane9991@gmail.com).
It bundles FastCtx (Apache‑2.0) – a separate project for fast file‑context handling.
Acknowledges the linux.do community for feedback.

TL;DR

Codey is a feature‑rich desktop manager for the Codex AI coding assistant, letting you juggle multiple accounts, switch between various LLM providers, monitor usage, and extend functionality with plug‑ins and light desktop automation. It’s aimed at power users who need more control than the vanilla Codex client offers.